Public Health & Ethics

That day, three patients arrived at a primary-care clinic at once: one had just returned from Southeast Asia with fever and myalgia, one had worked at an electroplating plant for five years with recurrent nasal congestion and nosebleeds, and the third was a community elementary-school teacher worried that gastroenteritis kept spreading through her class. Three seemingly unrelated threads — yet break each one down into which route the pathogen travels, who blocks it, and who lets it slip through, and the answers surface one by one along the very same causal chain. The chain of infection is a relay race: the pathogen leaves the reservoir, finds an exit, travels some mode of transmission, and reaches a susceptible host, and the essence of disease control is asking at which leg of the relay it is most efficient to knock the baton away.

A surgical mask mainly blocks droplets, so it works against droplet-predominant infections such as influenza and COVID-19, while airborne diseases such as tuberculosis and measles require an N95; but blood and body fluids travel by needlestick and fluid contact, vectors travel by mosquito bite, and food and water travel by ingestion — a mask sits on none of these three routes, and no number of layers will help. Waterborne protozoa hide an elegant trap: Giardia and Cryptosporidium survive as cysts, and cysts are chlorine-resistant — routine chlorination kills only bacteria and viruses, and the cysts survive regardless; only filtration can stop them. So that teacher's chlorinated water dispenser cannot stop Giardia; what actually works is sand filtration. Dengue works the same way — Taiwan's principal vector is Aedes aegypti, which prefers indoor standing-water containers, so vases, saucers, and buckets are the real breeding sites, while Aedes albopictus prefers the outdoors; so clearing standing water at home is not just a slogan, it cuts off the vector's favorite breeding spot. The surveillance pyramid is the easiest to get backward on exams — the base holds the number infected, including asymptomatic cases, the largest group, and the apex holds laboratory-confirmed cases, the smallest, with fewer people the higher up the pyramid you go. The incubation period tracks symptoms, the latent period tracks infectiousness, the serial interval tracks two people, and an asymptomatic person can already be infectious once the latent period ends but before symptoms appear — which is precisely the core reason asymptomatic transmission is so hard to contain.

The basic reproduction number is the average number of people one case infects with no immunity and no intervention at all — greater than one and the epidemic grows, equal to one and it holds steady, less than one and it recedes; once a vaccine or intervention is introduced, the actual transmissibility becomes the effective reproduction number, and the goal of disease control is to push it below one. The herd-immunity threshold equals one minus one divided by the basic reproduction number, so the larger the basic reproduction number, the higher the coverage required — measles has a basic reproduction number of roughly twelve to eighteen, pushing the threshold to ninety-two to ninety-five percent, and any drop in vaccination rate opens a breach. Exam questions love to conflate the basic reproduction number with the effective reproduction number; keep straight that the former is intrinsic to the disease while the latter reflects the moment after intervention. Taiwan's threshold numbers must be memorized precisely: foodborne illness is two or more people eating the same food with the same symptoms, but botulism or chemical poisoning can be confirmed from a single case; the HIV/AIDS elimination target has been upgraded from the old ninety-ninety-ninety to ninety-five-ninety-five-ninety-five, meaning ninety-five percent know their status, then ninety-five percent of those receive treatment, then ninety-five percent of those treated achieve viral suppression; notifiable communicable diseases are divided into five categories, and Category I must be reported within twenty-four hours. Vaccine policy is likewise never simply "vaccinate everyone" — prevalence, benefit, cost, and adverse effects must all be weighed; universal yellow fever vaccination in a non-endemic area should not happen, since it would only add adverse reactions.

The core of occupational exposure is the target organ: methylmercury causes the cerebellar ataxia and intention tremor of Minamata disease; lead causes anemia, abdominal colic, and wrist drop; cadmium causes the renal tubular and cartilage damage of Itai-itai disease; hexavalent chromium causes nasal septal perforation; arsenic causes blackfoot disease and Mees' lines. That electroplating worker's nosebleed is the signature perforation of hexavalent chromium. Pneumoconiosis is determined by which mineral dust is inhaled: silicon dioxide gives silicosis, asbestos gives asbestosis plus mesothelioma plus lung cancer, coal dust gives coal workers' pneumoconiosis; newspaper press workers inhale paper-fiber dust and do not belong to the mineral-pneumoconiosis high-risk group — this decoy comes up often. In the time-weighted average formula, the numerator is concentration multiplied by time, summed; the denominator is where most people get tripped up — it must be the sum of the actual time segments, not the statutory eight hours, so working only six hours means dividing by six, or the concentration gets underestimated. The greenhouse gases regulated under the Kyoto Protocol include carbon dioxide, methane, nitrous oxide, plus three groups of fluorinated gases; ammonia is an air pollutant but does not count as a greenhouse gas, and this is frequently picked by mistake. The key to food-safety management is the critical control point — the step "chicken must be cooked thoroughly" is a critical control point, because once control is lost there, the hazard cannot be removed downstream; acrylamide comes from the Maillard reaction in starchy foods during high-temperature frying or baking, with French fries and potato chips as the classic examples, classified as IARC Group 2A, possibly carcinogenic. Hold onto one pathway mindset for the whole chapter — masks, filtration, mosquito control, thorough cooking, gloves each cut a different link — and no matter how the question jumps, you can always locate it.

Three people came, one after another, into the exam room to sort out their lives: a man wanting to quit smoking, a middle-aged woman with multi-system discomfort after changing jobs, and an old man being referred to a major hospital with family accompanying him — none of them was asking what disease they had; they were asking what to do next. To answer this type of question, first put the theories back into three levels: individual, interpersonal, community. The Health Belief Model and the Transtheoretical Model both belong to the individual level, Social Cognitive Theory belongs to the interpersonal level, and the ecological model belongs to the community level. The spirit of Social Cognitive Theory is that person, behavior, and environment mutually influence one another, called reciprocal determinism; since it emphasizes interaction, it is not purely individual but interpersonal, and that is where it belongs. Its most important construct is self-efficacy, meaning confidence in one's own ability to complete a given behavior — don't confuse it with the Health Belief Model's perceived susceptibility, perceived severity, perceived benefits, perceived barriers, and cues to action, which are that model's own constructs.

The spirit of the Transtheoretical Model's five stages is one sentence: the right method for the right stage. Precontemplation doesn't know there is a problem, so consciousness raising must come first — providing information, raising awareness — and teaching smoking-cessation techniques directly at this point is wasted effort. Contemplation knows there is a problem and is willing but has not yet acted, calling for dramatic relief and self-reevaluation. Preparation already has a concrete action, such as signing up or buying equipment. Action is regular practice for less than six months. Maintenance is regular practice for more than six months, guarding against relapse. The difference between contemplation and preparation lies in whether a concrete action has been taken; that man who brought in an e-cigarette has already crossed from contemplation into preparation. The Health Belief Model, in turn, explains whether a behavior is adopted using the comparison of perceived threat against benefits and barriers; self-efficacy was only incorporated later, so it is shared with Social Cognitive Theory — this is the bridge point between the two schools.

That 50-year-old woman who changed jobs cannot be answered by any single specialty, because her story is exactly what the biopsychosocial model is meant to integrate. The biological layer is not just organ function and symptoms but also includes environmental factors that affect physiology; the psychological layer is affect, volition, personality, and coping style; the social layer is social support, culture, role, family, economics, and job change. When multi-system physical symptoms, emotion, and a social stressor appear together and the course stretches over weeks to months, the integrative answer is this model — don't pick just one organ. Family assessment often tests the Family APGAR, whose theoretical foundation is systems theory, treating the family as a system of mutually influencing parts; its five domains are adaptation, partnership, growth, affection, and resolve — don't confuse it with the neonatal Apgar score, which sounds alike but has entirely different content. Family medicine's six core attributes are the main course of reverse-inference questions: handling multiple problems in one visit is comprehensiveness; linking different levels and specialties through referral and consultation is coordination; following the same person long-term is continuity; an easily reached point of entry is first-contact accessibility; family-oriented means placing the person within the family context; person-centered means seeing the person, not just the disease. That old man being referred by his family doctor to a major hospital is most often mistakenly chosen as continuity, but the essence of a referral is linkage between institutions, so it belongs to coordination; comprehensiveness refers to handling multiple things in the same visit — the two are the pair most easily swapped.

The exercise prescription for chronic disease uses four letters: frequency, intensity, time, type. The moderate-intensity target heart rate is calculated as 220 minus age, then multiplied by 60 to 70 percent, so a 70-year-old's target is 90 to 105; applying 150 beats to a 70-year-old equals maximum heart rate and is dangerous. Exercise volume can also be accumulated in bouts rather than completed in one continuous session, which is friendlier to those with poor fitness or pain. The Narcotics Hazard Prevention Act provides that a first-time offender who merely uses a Schedule I or II substance may be sent to observation or mandatory rehabilitation rather than automatically prosecuted; prosecution targets those who manufacture, transport, sell, or transfer. Cultivation theory holds that heavy, long-term media viewing gradually leads people to believe the world in the media is the real world, and excessive violent content cultivates mean world syndrome; the Healthy City is a process, not an outcome, with cross-sector collaboration to continuously improve the environment at its core. Nonverbal communication carries roughly sixty to eighty percent of a message, not one-tenth; crossed arms signal defensiveness, not relaxation, while a slight head tilt plus a steady gaze signals interest. The three thresholds of Hall's four interpersonal-distance zones must be memorized: forty-five, one hundred twenty, three hundred sixty — and a physical examination crosses into the intimate zone. The core of the whole chapter is putting the person back in the exam room: not just looking at organs, but hearing the sections of behavior, family, and society as well.

At the monthly research meeting, an attending announces a beautiful research conclusion — counties with higher cigarette sales have higher incidence of coronary heart disease, so smoking causes coronary heart disease — and a senior colleague immediately points out that this is county-versus-county, not person-versus-person, called the ecological fallacy. To avoid this trap, look at the direction of the timeline first and everything falls into place. Grouping by exposure and following forward for disease is the cohort study, which can calculate incidence and relative risk and can establish temporality — its greatest strength. Grouping by disease presence or absence and looking back for exposure history is the case-control study, which can only calculate the odds ratio, though under the rare-disease assumption the odds ratio approximates the relative risk, while for a common disease the odds ratio overestimates it. Measuring at the same time point is cross-sectional, which can only calculate prevalence, with an unclear causal timeline. Using the group as the unit of analysis is the ecological study, which is prone to the ecological fallacy. Running a case-control study inside an existing cohort is called a nested design — a bidirectional design in which biological samples are banked in advance, reducing recall bias. The randomized controlled trial uses random allocation to simultaneously balance known and unknown confounders, and sits at the highest level of causal evidence.

Confounding and effect modification are opposite in nature. Confounding is a third variable that affects both exposure and outcome at once, producing a spurious association, and must be controlled or removed — matching, stratification, or multivariable adjustment can be used; effect modification is a real difference in association strength across different levels, and must be reported by stratum, never eliminated. For example, age being associated with both exercise and heart disease is confounding, while a given genotype having a different drug response than other genotypes is effect modification. The true purpose of matching is to control confounding and improve comparability between the two groups — not to increase heterogeneity, and not for generalizability or to lower the non-response rate. Each design's characteristic bias must also be matched correctly: the case-control study fears recall bias, because cases try harder than controls to recall past exposure; the cohort fears loss-to-follow-up bias, because over a long follow-up period, dropouts differ from those who remain; screening-related studies fear lead-time bias, because early diagnosis merely shifts survival time earlier without truly extending life.

Choosing a test doesn't require rote memorization — three questions suffice: is the dependent variable continuous or categorical, how many groups, and is the data independent or paired. Two continuous, independent groups use the independent-samples t-test; two continuous, paired groups use the paired t-test; three or more continuous groups compared by mean use analysis of variance, because switching to repeated pairwise t-tests would inflate the Type I error rate. Use the Z-test when the population SD is known or the sample is large; use chi-square for two unpaired categorical variables; a 1:1 matched binary categorical variable must use McNemar's test rather than the independent chi-square test, because each pair is one unit and the two observations are not independent — only the discordant pairs provide matching information, while concordant pairs provide none; when any cell in a 2×2 table has an expected count under five, switch to Fisher's exact test, because the chi-square approximation has already broken down. A continuous dependent variable influenced by multiple independent variables uses linear regression. The confidence interval and hypothesis testing are, in fact, equivalent: the null value for a difference is zero, the null value for a ratio is one, and if the confidence interval covers the null value, it is not significant — so that interval running from 0.028 to 0.202, which covers 0.15, must be judged as not statistically significant; an odds-ratio interval from 0.7 to 1.5 covering 1 is likewise not significant, while a relative risk from 1.2 to 2.4 that excludes 1 is significant.

A decision can commit two opposite errors: Type I is a false positive conjured from nothing, with probability called α; Type II is a false negative that misses the truth, with probability called β. Statistical power is one minus β, meaning the probability of correctly detecting a difference when one truly exists; the most direct way to raise power is to increase the sample size, so the most common consequence of an inadequate sample is insufficient power, making a Type II error likely — judging a real difference as not significant. Error likewise falls into two families: random error has no fixed direction, lowers precision, and can be canceled out by averaging over repeated measures — the blood-pressure fluctuation caused by a cuff that deflates sometimes fast, sometimes slow, is an example; systematic error deviates in a fixed direction, lowers accuracy, and averaging cannot remove it — a sphygmomanometer that was never zeroed and reads five high every time is an example. In one line: random error is erratic, systematic error is skewed. For central tendency in a skewed distribution, use the median rather than the mean, because the median is unaffected by extreme values; use the standard deviation to describe the spread of the data, since the standard error describes the uncertainty of the sample mean, not the spread of the data itself. Once age-stratified rates have been calculated for each stratum, two different populations must not simply be pooled into one overall rate — doing so introduces confounding bias, and standardization must be used instead. The last easy point is Mendelian genetics: a dominant heterozygote Aa crossed with a normal aa gives each pregnancy a one-half chance of being affected, and both of two children being affected requires multiplying to one-quarter. Hold onto the two axes of timeline and hypothesis testing and the whole chapter falls into place.

In the clinic, a physician complains that Group A has so many false positives — is the tool broken? The researcher shakes his head: the tool isn't broken, Group A's prevalence is simply too low. The root of every metric is the 2×2 table: those truly diseased split into test-positive and test-negative, and those truly disease-free split the same way. Sensitivity is the proportion of the truly diseased who are caught; specificity is the proportion of the truly disease-free who are excluded; positive predictive value is the proportion truly diseased among those testing positive; negative predictive value is the proportion truly disease-free among those testing negative. If the negative predictive value is eighty percent, then twenty percent of those testing negative are still truly diseased — that is the fish that slips through the net.

Sensitivity and specificity are the difficulty of the exam paper, determined by the test's inherent biological properties and threshold, and do not change with prevalence; positive and negative predictive values are this particular cohort's scores, varying with the prevalence in the source population. When prevalence is high, a positive result is more trustworthy; when prevalence is low, the proportion of false positives among all positives balloons and the positive predictive value plunges — so Group A's high false discovery rate is correctly explained by low prevalence, not by the quality of the tool. The likelihood ratio, like sensitivity and specificity, is unaffected by prevalence: the positive likelihood ratio is sensitivity divided by one minus specificity, and above ten it strongly confirms a diagnosis; the negative likelihood ratio is one minus sensitivity divided by specificity, and below 0.1 it strongly rules one out. What matters most to clinical practice is the positive and negative predictive value, because they answer what this particular positive result means for me.

Threshold and test combination form a symmetric seesaw. Raising the threshold makes the standard stricter — the false positive rate and sensitivity both fall while specificity rises, so fewer innocent people are wrongly accused but more guilty ones slip away; lowering the threshold does the reverse. The receiver operating characteristic curve plots sensitivity on the vertical axis and one minus specificity on the horizontal axis; the closer the area under the curve is to one, the better. Combination testing is likewise symmetric: parallel testing calls any one positive a positive — catching more, releasing fewer, so sensitivity rises and specificity falls — and is used in critical illness where missing a diagnosis is feared; serial testing requires every test to be positive, guarding the gate at every layer, so specificity rises and sensitivity falls, and is used to confirm a diagnosis or reduce expensive invasive testing. Treatment benefit should be judged by absolute risk reduction and the number needed to treat; looking only at relative risk reduction invites the "relative" framing to mislead you — a fifty percent relative risk reduction might just be a drop from two to one, an absolute reduction of only one, with a number needed to treat of one hundred. Screening evaluation carries three biases that falsely inflate survival: lead-time bias merely shifts the diagnosis time point earlier while the time of death is unchanged; length-time bias more easily catches slow-progressing, good-prognosis cases; and overdiagnosis detects lesions that would never have caused disease — so evaluating the effectiveness of screening must look at disease-specific mortality rather than survival, or the biases will inflate the numbers.

Screening belongs to the third stage of secondary prevention — the same fact asked two ways: ask for the level and the answer is secondary, ask for the stage out of five and the answer is the third — don't be misled. And the core of the WHO's ten principles of screening is that the disease must be treatable, and early detection must be able to change the outcome early, or it merely manufactures anxiety. As for the direction of screening decisions: low-prevalence mass screening should choose high specificity, because false positives are the main problem; fearing to miss a severe but treatable disease calls for high sensitivity; ruling out disease uses a high-sensitivity negative result, ruling in disease uses a high-specificity positive result. Among levels of evidence, a meta-analysis is Level I, corresponding to a Grade A recommendation; if an option labels it II-A, that's a contradiction. In building a PICO question, the intervention field must name the intervention actually under study, not some other intervention or the outcome. Two must-know USPSTF Grade A items: women planning pregnancy should take 0.4 to 0.8 mg of folate daily to prevent neural tube defects, and sexually active individuals need intensive counseling for STI prevention (actually a Grade B item, renamed behavioral counseling in 2020) — brief counseling isn't enough, because changing sexual behavior involves complex psychosocial relationships. The vaccine-age trap is a frequent favorite: a 15-year-old girl does not need zoster vaccination, because it is meant for immunocompetent people aged fifty and above; the HPV vaccine is the one whose main target is ages nine to twenty-six. Hold onto the 2×2 table as the root of the whole chapter, reason along the two lines of a test's intrinsic properties versus differences in the source population, and no matter how the question twists, you won't get dizzy.

Mid-consultation, an older physician sighs that fee-for-service drives up volume, per-case payment controls cost, and capitation is the cheapest but most prone to corner-cutting — the payment system itself decides how we see our patients. The genetic code of Taiwan's National Health Insurance is mandatory enrollment plus a single government-run payer; the National Health Insurance Administration under the Ministry of Health and Welfare is the sole purchaser, negotiating with every healthcare provider. The size of the settlement unit, from small to large, determines the direction of the incentive: the larger the unit, the stronger the incentive to economize, but also the greater the risk of sacrificing service volume — so fee-for-service drives up volume, per-diem payment may prolong admissions, per-case payment may prompt early discharge and cherry-picking of mild cases, and capitation is the cheapest but most prone to corner-cutting. The global budget payment system in place since 2002 is overall supply-side cost control.

Information asymmetry in the healthcare market makes the physician simultaneously agent and supplier; if this advantage is exploited to induce excess consumption, that is supplier-induced demand, belonging to the supply side. Moral hazard is the patient's overuse of care after obtaining insurance, because out-of-pocket cost is low — it belongs to the demand side. Countermeasures likewise split into two sides: the supply side uses global budgets, per-case payment, clinical guidelines, and second opinions; the demand side uses co-payment. Among management tools, in SWOT the internal factors are strengths and weaknesses, the external are opportunities and threats — the O is often misremembered as "objective." In the Balanced Scorecard, finance is a lagging indicator reflecting past results, while learning-and-growth and internal process are leading indicators predicting the future. The WHO's Healthy City health-category indicators do not include the abortion rate (it is a socioeconomic indicator), and the Pharmaceutical Affairs Act's drugs do include medical devices — these two facts are often buried as false options.

The four branches of shock are sorted by four hemodynamic parameters: hypovolemic has low preload, high afterload, low cardiac output; cardiogenic has high preload, high afterload, low cardiac output; obstructive has high preload, high afterload, low cardiac output — these first three are called cold shock. Distributive shock is warm shock, with afterload falling while cardiac output rises, and its representatives are sepsis, anaphylaxis, and neurogenic causes. For a quick sort, look at the jugular veins: flat jugular veins point to hypovolemic, distended veins point to cardiogenic or obstructive, and cardiac tamponade is read from Beck's triad — hypotension plus jugular venous distension plus muffled heart sounds. The order of first-line management must be second nature: hypovolemic gets large-volume crystalloid; septic gets crystalloid first, then a vasopressor with norepinephrine as the choice rather than dopamine; anaphylactic shock's first line is intramuscular epinephrine 1:1000 in the lateral thigh, not antihistamines or steroids given first; tension pneumothorax gets needle decompression first, then a chest tube; cardiac tamponade gets pericardiocentesis.

The Parkland formula for burns gives a total volume equal to four milliliters times body weight times the percentage area of second- plus third-degree burns — first-degree does not count — with half given over the first eight hours counted from the time of injury, not the time of arrival, titrated to urine output. The most common mistake is adding in the first-degree thirty percent as well, arriving at seventy percent and 7,000 mL of fluid; the correct calculation counts only the second- plus third-degree forty percent, giving 8,000 mL, with 4,000 mL over the first eight hours. The rule of nines for adults is head nine, each upper limb nine, trunk front and back eighteen each, each lower limb eighteen, perineum one; in children the head accounts for a larger share, and the Lund-Browder chart is more accurate. As for blood alcohol concentration, 0.30 brings light coma and unstable vital signs, and only above 0.40 does respiratory depression become fatal. The mechanism of referred pain is that visceral afferents and somatic sensory afferents converge at the same spinal segment, and the brain cannot tell them apart, misreading it as somatic pain — so cardiac pain is referred to the medial left arm and jaw, gallbladder pain to the right shoulder, pancreatic pain to the mid-back, early appendiceal pain to the periumbilical area before shifting to the right lower quadrant, and ureteral stone pain to the groin; using the reverse direction — pressing on the skin to evoke or localize the visceral source — is a mistaken application.

Headaches divide into primary and secondary: the most common primary type is tension-type, bilateral, pressing, non-pulsatile; migraine is unilateral and pulsatile with nausea and photophobia, and may carry an aura; cluster headache is severe periorbital pain in men with tearing and nasal congestion. The five-point red-flag mnemonic is systemic symptoms, neurologic abnormality, thunderclap onset, new onset in an older patient, and a change in pattern; a thunderclap headache must first rule out subarachnoid hemorrhage with a CT scan, with lumbar puncture if needed. Medication-overuse headache is one that worsens the more you treat it, and the management is withdrawal, not a higher dose. The core of lymphedema care is complete decongestive therapy; emollients actually protect the skin and lower the risk of cellulitis rather than being contraindicated, and the affected limb should avoid blood pressure measurement, blood draws, injections, overheating, and constriction. On disaster timing, typhoons bring relatively few casualties in the acute phase and more in the recovery phase, because drowning, landslides, cleanup injuries, and generator carbon-monoxide poisoning all come afterward; earthquakes are the opposite, with most casualties concentrated within hours of the event from collapse and crushing, and crush syndrome afterward brings rhabdomyolysis, hyperkalemia, and myoglobinuria causing acute kidney injury that requires aggressive fluid resuscitation. The three zones of a chemical disaster are: the hot zone only rescues and extracts casualties, the warm zone performs decontamination, and the cold zone performs triage and treatment — placing decontamination in the cold zone is wrong. Read the whole chapter through incentives, hemodynamics, and formulas, and follow the mechanism, and the answer will present itself.

Three sets of patients pass through the family medicine clinic one after another: a demented old woman grows agitated after surgery, a frail old man has lost five kilograms, and the family of a patient with terminal lung cancer asks whether he can be made more comfortable. Delirium turns on two axes: predisposing factors are the patient's own baseline vulnerability, and precipitating factors are the acute external insult. Dementia is the single most important predisposing factor for delirium in hospitalized older adults, because cerebral reserve is already low and almost any small insult can set it off; common precipitants include infection, surgery, dehydration, electrolyte disturbance, anticholinergics, opioids, benzodiazepines, and physical restraint. The incidence of delirium in hospitalized older adults runs as high as fourteen to fifty-six percent, not one to five percent. Delirium differs from dementia: delirium is acute, fluctuating, marked by impaired attention, and reversible, while dementia is chronic, progressive, predominantly a memory disorder, and usually irreversible; the interventions that actually worsen delirium are restraint, benzodiazepines, anticholinergics, and opioids, so these aggravate it rather than treat it.

Most parameters decline with aging: maximal heart rate is derived from two hundred twenty minus age, and arterial oxygen tension, glomerular filtration rate, muscle mass, and basal metabolic rate all fall — but there is one elegant exception, serum insulin concentration, because aging commonly brings rising insulin resistance, so insulin may hold steady or even rise rather than fall, which is why it is so often the answer when a question lists a string of parameters that decline with age. The instruments for functional assessment in older adults must be kept straight: the MMSE assesses cognition, not IQ or communication; the Barthel Index covers basic activities of daily living such as feeding, toileting, transferring, bathing, and dressing, and excludes self-medication; self-medication — sorting a pillbox, tracking the time — belongs to the instrumental activities of daily living; the TUG assesses gait and fall risk, with normal being under twelve seconds and over twelve to fourteen seconds signaling high fall risk. The five Fried frailty criteria are weight loss, exhaustion, weakness, slow gait speed, and low activity level; meeting three or more defines frailty, one to two defines pre-frailty, and the key trap is that the five criteria exclude cognitive function — cognitive frailty is a separate dimension. High risk for hypothermia includes extremes of age, hypothyroidism, shock, malnutrition, and alcohol, but not healthy adults aged thirty to fifty; for altered mental status in an older adult, think sugar, flow, electrolytes, infection, drugs — hypolipidemia causing confusion is a distractor; high risk for malnutrition includes critical illness, old age, alcoholism, chronic disease, malignancy, and malabsorption, while drinking soda runs the opposite direction as empty calories and is not a risk factor.

The core of palliative and hospice care is shifting the goal from cure and life prolongation to symptom relief and a good death; the WHO defines it as neither hastening nor postponing death, so an option reading "prolongs survival time" is wrong. End-of-life symptom management follows a logic from mechanism to first-line drug: dyspnea gets low-dose morphine because it lowers the perception of ventilatory drive and works even when oxygen saturation is normal, though reversible causes such as bronchospasm, pleural effusion, and cardiac tamponade must first be ruled out; terminal delirium gets haloperidol because it blocks dopamine with low sedative burden, not a benzodiazepine, which would make it worse; cancer pain follows the WHO analgesic ladder from non-opioid to weak opioid to strong opioid, dosed on a fixed schedule with a rescue dose for breakthrough pain, and severe pain goes straight to a strong opioid; intestinal colic and secretions are treated with hyoscine. Supplemental oxygen only benefits dyspnea from hypoxemia — when oxygen saturation is normal, a fan blowing across the face relieves it instead; whenever an opioid is prescribed, a laxative must be co-prescribed as routine, because opioid-induced constipation does not develop tolerance even though nausea does. Artificial nutrition and hydration in the dying phase cannot improve survival or comfort for an obtunded patient dying of cancer, and instead increase edema, ascites, aspiration pneumonia, and secretions; thirst is more often relieved with oral care and lip moistening.

The two statutes must be kept distinct in scope: the Hospice Palliative Care Act covers only terminally ill patients and allows refusal of CPR and life-sustaining treatment, while the Patient Right to Autonomy Act extends to five categories — terminal illness, irreversible coma, permanent vegetative state, extremely severe dementia, and other officially announced conditions — and further allows refusal of life-sustaining treatment and artificial nutrition and hydration, with the core document being the advance directive, which requires advance care planning beforehand; an option stating that dementia of any severity qualifies is wrong, since it must be extremely severe. The priority order for DNR decisions places the patient's own wishes highest; once the patient loses capacity, the question becomes whether a pre-signed declaration of intent or an appointed healthcare proxy exists, and a proxy's legal authority takes priority over the generic order of relatives, with a consent form from the nearest relative used only if neither exists. The statutory order for relatives signing a DNR on the patient's behalf is spouse, adult children and grandchildren, parents, siblings, and grandparents — remembered as spouse, children and grandchildren, parents, siblings, grandparents — and the common way to lose points is ranking parents ahead of adult children, when adult children in fact take priority. The terms must also be kept distinct: when the patient signs it personally it is called a declaration of intent, and when a relative signs on the patient's behalf it is called a consent form. Among the four practices at the boundary of hastening or not prolonging death, palliative and hospice care is legal, and withdrawing life support is legal because it permits natural death, while euthanasia and physician-assisted suicide are illegal in Taiwan — the difference lies in who acts and whether the intent is relief or the deliberate causing of death; the WMA's 2019 declaration opposes euthanasia and physician-assisted suicide but supports a physician's duty to respect a patient's informed refusal of life-sustaining treatment, because dying as a result of refusing treatment is an exercise of autonomy, not euthanasia. For a newborn with a severe congenital anomaly such as Edwards syndrome, when the parents, after being fully informed, choose against aggressive invasive treatment, palliative care should be provided — this is not the same as abandonment. Home hospice care offers a familiar environment, companionship, and dignity, but its symptom control is less convenient than on an inpatient ward. Hold onto the chapter's two main threads — triggers and goals — and every bedside decision falls into place.

In a research meeting, an IRB member checks a consent form line by line for whether it states the alternative treatments and the right to withdraw at any time; at an ethics consultation, a family member stops the attending physician and begs him not to tell the patient; at night, a medical intern flips through a neighbor's chart purely out of curiosity, telling himself it should be fine since he hasn't told anyone else. Three scenes, three statutes, three frameworks of principle — get any of them wrong, and the price is not a lost exam point but a patient's rights. The four principles proposed by Beauchamp and Childress are autonomy, nonmaleficence, beneficence, and justice, and they hold that all four carry prima facie status, meaning they are equally weighted with no fixed order of priority, and when they conflict, the resolution must be worked out contextually, so an option stating that autonomy always ranks highest or that the four have a fixed order is wrong. Nonmaleficence means not causing harm, beneficence means actively pursuing benefit, and the two point in different directions and must not be swapped; justice emphasizes fairness in the distribution of resources and treatment, and must not be omitted or mistaken for mere legality.

A family asking that the patient not be told he has cancer is a classic test scenario: the physician should decline on the basis of respecting autonomy, because the subject of disclosure is the patient himself, and the family's good intentions cannot substitute for the patient's right to know his own condition — treating the family's wishes as beneficence, or treating "disclosure would upset the patient" as nonmaleficence, are both misapplications. The rare exception is therapeutic privilege, which may only be invoked under the high threshold that disclosure would very likely cause serious, immediate physical or psychological harm, and requires substantial clinical justification — it cannot be invoked merely because the family asks or fears the patient will be upset. But if a patient with decision-making capacity actively states, "I don't want to know — please tell my family," the physician may inform the family under that authorization, and this remains an exercise of the autonomy principle, not the family overriding the patient. Valid informed consent requires all five elements together — decision-making capacity, adequate disclosure, understanding, voluntariness, and consent/authorization — and adequate disclosure must include the diagnosis, the recommended treatment, its benefits and risks, the alternatives, and the consequences of forgoing treatment, with alternatives being the element most often omitted; coercion violates voluntariness, inadequate information violates adequate disclosure, and being unconscious, a minor, or having severe dementia reflects a lack of decision-making capacity — do not conflate the three.

The three cornerstones of research ethics: the Nuremberg Code of 1947 emphasizes that voluntary informed consent is absolutely essential and is the starting point; the Declaration of Helsinki of 1964, drafted by the World Medical Association, holds that the welfare of research subjects always takes priority over the interests of science and society; the Belmont Report of 1979 proposes three principles — respect for persons, beneficence, and justice — and forms the theoretical foundation of the U.S. IRB system; the Declaration of Lisbon of 1981 is a declaration on patient rights, not a code of research ethics, and must not be confused with the others. The iron rule is that the rights, safety, and welfare of research subjects always take priority over their potential contribution to science and society. Helsinki specifically stipulates that under a dependent relationship, informed consent must be obtained by a qualified individual who thoroughly understands the research but is completely independent of the physician-patient relationship, and an impartial witness who merely watches the patient sign is not sufficient. Vulnerable populations include children, pregnant women, fetuses, embryos, prisoners, the terminally or critically ill, the cognitively impaired, the mentally ill, and the economically or educationally disadvantaged; a healthy older adult alone does not belong to the core vulnerable populations traditionally enumerated, while the embryo does. Authorship is listed according to actual contribution, not position; a guest author violates authorship ethics, and the corresponding author should be someone who genuinely participated and can be held accountable for the content. Clinical trial records are retained permanently — even if the subject is a seven-year-old child, the record is still kept permanently rather than falling under the "seven years after majority" rule, because the trial requires long-term tracking of safety and protection of the subject.

Public health ethics is another axis: mandatory public health policies such as quarantine, compulsory vaccination, or smoking bans must satisfy the principle of proportionality — the means must be conducive to achieving the goal and must adopt the least restrictive approach, and cannot be enforced without exception, since considerations such as bodily integrity and religious belief must still be weighed as possible exceptions; smoking is a matter of personal autonomy, but secondhand smoke harms others and already exceeds the scope of pure autonomy, constituting a public health ethics issue, so the state may intervene with restrictions such as smoking bans in public venues — in one sentence, your freedom ends at the tip of someone else's nose. Medical harm is divided into fault-based and no-fault harm; a known surgical complication, anesthesia risk, or an unpredictable drug allergy is no-fault harm, connecting to the Medical Incident Prevention and Dispute Resolution Act, which emphasizes compassionate response, mediation, and error correction. In clinical trials, Phase I tests safety and dosing, Phase II tests preliminary efficacy and dose-finding with later stages often including a control, Phase III is a large-scale randomized controlled trial confirming efficacy and comparing against the standard treatment, and Phase IV is post-marketing surveillance; an option stating that Phase II tests only safety and involves neither efficacy nor a control is wrong.

Confidentiality is the default principle but has statutory exceptions: statutorily notifiable communicable diseases must be reported; child and adolescent abuse, sexual exploitation, and sexual assault of a person under eighteen, as well as domestic violence, require mandatory reporting; suspected sexual assault, regardless of the victim's age, must be reported to the competent authority; under the spirit of Tarasoff, clear and serious danger to an identifiable third party may or must be warned of and disclosed; and a court subpoena or judicial investigation must be complied with as required by law. What must be kept distinct is that mandatory reporting to the competent authority is a statutory duty of healthcare workers and does not depend on the victim's consent or on whether the victim reports to the police, even in cases of adult sexual assault, whereas reporting to the police and undergoing forensic evidence collection for referral to criminal investigation should respect the adult victim's own wishes; the reporting deadline is immediate, no later than twenty-four hours, and the identity of the reporter must be kept confidential. Older study notes commonly state that adult sexual assault requires the victim's consent before it can be reported, but this actually conflates mandatory administrative reporting with criminal reporting and evidence collection — under current law, reporting to the competent authority is mandatory. Confidentiality does not conflict with truthful charting: a physician may promise not to disclose information externally, but the medical record must be charted truthfully and completely under the Medical Care Act and may not be omitted or falsified at the patient's request — the former is permissible, the latter is unlawful. Custody of medical record documents belongs to the healthcare institution, for at least seven years, until seven years after majority for minors, and permanently for human trials; the patient has the right to know, obtain copies of, and have the personal data in their own medical record protected, and the paper record belongs to the hospital, though the patient may request copies. Clerkships at teaching hospitals require prior notice and consent; accessing an unrelated record without authorization violates confidentiality and personal data protection even if never disclosed externally, since access to a medical record requires a legitimate, work-related need to know. Medical Care Act Article 106, the offense of obstructing medical practice, is a non-complaint public offense, and the prosecutor may investigate on their own initiative without waiting for the victim to file a complaint; Medical Care Act Article 79 requires that informed consent for human trials specify the available alternative treatments and the right to withdraw consent at any time. This chapter converges on a single thread: public health is the path, statistics is the tool, and ethics is the boundary — once you understand why, the test points grow on their own.

At two in the morning the lab spits out a report, potassium seven point two, and that patient's attending has just clocked out, with the handoff sheet reading only "stable condition, nothing pending." Whether that call was made, who it was made to, whether the recipient read it back, and whether it was written into the chart — these decide whether that old man is still alive tomorrow morning. The 1999 Institute of Medicine report, *To Err Is Human*, put forward an estimate that no one wanted to accept at the time: tens of thousands of people die every year from preventable medical errors, and harder still to accept was its conclusion — these errors are not caused by bad physicians, they are caused by good physicians working inside bad systems. Human attention, memory, and communication all have physiological limits, and a hospital is precisely the place that pushes those limits to their extreme, so being a little more careful is not a solution that scales — it would require every person to be at their best every single day, which is statistically impossible. The Swiss cheese model describes exactly this: every layer of protection has holes, and an accident occurs the instant all the holes happen to line up; since the holes cannot be eliminated, the goal of systems engineering becomes stacking more layers and keeping the holes from aligning.

Let's start with critical values. An abnormal value is defined as outside the reference range, and it may simply be the long-standing baseline of a patient with chronic kidney disease; a critical value is defined without any regard to the magnitude of deviation, but entirely by time — without immediate action, the patient may die or suffer irreversible harm within hours. So what the critical-value system actually governs is not how accurate the test is, but whether the information reached the person with authority to act on it within that window of time, which is why the lab's two major quality metrics are both time-based, one being turnaround time and the other the critical-value notification interval. Potassium below two point five or above six point five directly shifts the myocardial resting membrane potential and causes lethal arrhythmia; glucose below fifty cuts off the brain's sole fuel source, and above five hundred heads toward ketoacidosis and hyperosmolar dehydration; sodium below one hundred twenty or above one hundred sixty causes brain cells to swell or shrink; calcium below six or above thirteen causes a drastic change in neuromuscular excitability; platelets below twenty thousand cause spontaneous bleeding, including intracranial hemorrhage; an international normalized ratio above five in a patient on anticoagulants signals a steep rise in major bleeding risk; a rising troponin means time is myocardium; a blood culture positive in even a single bottle means bacteremia, and every hour of delayed antibiotics raises mortality; bacteria seen on a cerebrospinal fluid smear means bacterial meningitis; an arterial blood gas pH below seven point two is severe acidemia. The exact thresholds are set by each hospital, and the exam tests the order of magnitude and the logic, not memorized numbers. The real test point is in the form: notification must be closed-loop, and all three elements are mandatory — notify the person who can personally give the order, have them read back the complete value and patient identity, and write the time, recipient, and response into the record; flagging it red, posting in a group chat, or leaving a message with a clerk all fail to count as delivery. The Joint Commission, in 2005, added the timely reporting of critical laboratory results to its National Patient Safety Goals and required that the recipient read the value back during a telephone report.

At the moment of a shift change, three dangerous things happen at once: information has to be reconstructed from one mind into another while the brain retains only a summary; responsibility transfers in that instant, yet the boundary of that transfer has never been declared; and the person taking over is, at this moment, the one who knows the least about this patient in the entire hospital, yet holds the greatest authority. What a handoff truly needs to transmit is not data but a mental model — which trajectory the patient is on, what is most likely to happen next, and what the first action should be if it does; raw data can be looked up in the electronic medical record, but a mental model cannot, so its value lies almost entirely in that latter half, prediction and contingency. SBAR solves the problem of the speaker being unable to articulate clearly — situation, background, assessment, and recommendation — and its power lies in forcing the last two fields to be filled in, which are exactly the two fields a junior team member is most afraid to voice; the essence of SBAR is giving someone lower in the hierarchy a legitimate slot to state a judgment out loud, and it originated in the communication format used on U.S. Navy nuclear submarines before being introduced into healthcare in the late 1990s. I-PASS solves the problem of the listener thinking they already understood — illness severity, patient summary, action list, situation awareness and contingency planning, and finally synthesis by receiver, which requires the person taking over to restate what they just heard in their own words, and that is the true heart of the whole protocol; a 2014 multicenter study from a Boston Children's Hospital team showed a significant drop in medical errors and preventable adverse events after implementation. You will notice this runs on the same principle as the read-back for critical values — a message only counts as delivered once it has been read back. Responsibility after a handoff does not disappear, it only splits into two segments: the person handing off is responsible for whether they spoke clearly, and the person taking over is responsible for whether they acted reasonably on the information received, and the only vehicle that can prove either is the medical record; Article 68, Paragraph 1 of the Medical Care Act requires personal documentation with a signature or stamp and the date, and Paragraph 3 stipulates that a verbal order must be followed by a written record within twenty-four hours.

Eight in the morning, Orthopedics Operating Room 3: the consent form says left knee, but the patient has been positioned for the right leg, and the circulating nurse opens her mouth and closes it again — in eleven years at this hospital, she has never once interrupted an attending physician before an incision. What she needs is not courage, but a legitimate moment to speak, granted to her by the system. Wrong-site surgery, wrong-patient surgery, and retained foreign objects are almost never a matter of no one knowing — in that preoperative room, someone always has the correct answer in hand, and the error occurs because that person never voices it, and the reason is hierarchy. So the checklist's true function is not helping everyone remember, but converting a default assumption into a spoken declaration — instead of everyone silently assuming everyone else already knows, someone must now say it aloud and someone must answer; more critically, it creates an institutionally sanctioned moment to speak, during which anyone speaking up during those thirty seconds of the pause is not an offense but simply following procedure, and team self-introduction removes the sense of anonymity so people find the courage to speak. The World Health Organization's 2008 Safe Surgery Saves Lives campaign introduced a nineteen-item checklist, led by Gawande, and an eight-country study in 2009 showed in-hospital mortality falling from one point five to zero point eight percent and major complications falling from eleven to seven percent. The three checkpoints must be kept distinct: sign in happens before anesthesia induction, while the patient is still awake, confirming identity and site himself and checking the anesthesia machine and pulse oximeter; the time-out happens before skin incision, with the whole team pausing, introducing themselves to each other, verbally confirming the patient, procedure, and side, and asking whether prophylactic antibiotics were given within sixty minutes; sign out happens before the patient leaves the room, confirming the name of the procedure actually performed, the instrument and sponge and needle counts, and the specimen labeling and name check. The site marking must be made by the operating surgeon with the patient awake and participating, marked at the incision site; the standard response to a discordant count is to recount, search, and obtain an intraoperative X-ray if needed, not to close the wound outright; wrong-site surgery, wrong-patient surgery, and retained foreign objects are never events. The biggest practical trap is letting the checklist degenerate into signing it after the fact — the active ingredient is verbal execution, spoken aloud and answered aloud, since a signature on paper is only a record, not an intervention.

Accrediting bodies sell neither drugs nor devices — what they sell is whether you are eligible to be paid. The Joint Commission was founded in the United States in 1951, and accreditation grants deemed status under federal Medicare; its international division was founded in 1998 to sell certification to hospitals worldwide; Taiwan's Joint Commission of Taiwan was founded in 1999, commissioned to conduct hospital accreditation and teaching-hospital accreditation, with results linked to hospital tiering, teaching status, resident training quotas, and National Health Insurance contracts. A published paper depends on every physician individually reading it, believing it, and changing their habits, a path of diffusion measured in decades; an accreditation standard needs no one to believe it, it only needs to be written into the rulebook, and because it is tied to the survival conditions of the entire hospital, it becomes nationwide routine within a single year. The cost is formalism — when a standard only checks whether there is a record, the most rational response for a hospital is to produce records rather than produce safety, so the quality of accreditation itself depends on whether it checks process or outcome. Another system running on opposite logic is the reporting system: Taiwan's Patient-safety Reporting system was built in 2003, on the principles of anonymity, voluntariness, confidentiality, non-punitiveness, and shared learning, because the moment there is accountability, no one reports and the system goes blind instantly; paired with it is the concept of the second victim — in an incident, the patient and family are the first victims and the healthcare worker involved is the second victim, and an institution that only assigns blame without offering support will inevitably end up with concealed reporting and loss of talent.

Finally, the two tubes. The four axes of healthcare-associated infection are ventilator-associated pneumonia, surgical site infection, central line-associated bloodstream infection, and catheter-associated urinary tract infection. A central line gives bacteria three routes — pushing skin flora in at the moment of insertion, migrating inward along the external surface of the catheter, and contaminating the internal lumen through the hub — so every element of the care bundle corresponds to one route: maximal sterile barrier precautions plus chlorhexidine antisepsis block the first, daily dressing inspection blocks the second, and disinfecting the hub before every use blocks the third; site selection is a trade-off, not a ranking — the femoral vein, close to the perineum and abraded by hip flexion, has the highest infection rate, while the subclavian vein has the lowest infection rate but a higher risk of pneumothorax; and the single most effective intervention is the least technical of all, assessing daily whether this line is still needed, because on the day there is no line, the infection rate is zero. Starting in 2003, a Michigan program used a five-step sheet of paper to drive ICU catheter infection rates down to near zero, often misread as proof the checklist worked, when the real variable was that it simultaneously authorized nurses to halt the procedure whenever a physician skipped any step. A urinary catheter turns an originally sterile, self-cleaning bladder into a fixed conduit, with bacteria ascending along the inner and outer walls and forming a biofilm, and bacteria within that biofilm are relatively immune to antibiotics and immune defenses, so duration of catheterization is the single strongest risk factor, with risk a function of time, and the only effective intervention is to shorten that time — do not insert one without an indication, and once inserted, ask every day whether it can come out; industry spent twenty years trying to sidestep this conclusion with antimicrobial-coated tubes, but large randomized trials showed the benefit insufficient to support routine use, and the correct answer is one that nobody can sell. Four traps must be remembered: asymptomatic bacteriuria is not treated, with the only exceptions being pregnant women and patients before a procedure likely to cause bleeding of the urinary tract mucosa; catheter-tip culture cannot diagnose infection, and the catheter must be replaced first with the specimen taken from the new one; routine catheter changes, bladder irrigation, and prophylactic antibiotics are all ineffective; and the drainage bag must stay below the bladder, off the floor, and sealed. Finally, there is money — starting October 1, 2008, the United States stopped providing additional payment for hospital-acquired conditions not present on admission but arising during the hospital stay, including central line-associated bloodstream infection and catheter-associated urinary tract infection; this rule mandates no clinical practice whatsoever, it simply shifts the cost of infection from the insurer back onto the hospital, turning infection control from a department that spends money into a department that protects revenue, which changed behavior faster than any published paper ever could.

An afternoon on the hospice ward. An eighty-two-year-old man with advanced dementia has not spoken in two years, and the nurse finds him during his afternoon nap with his brow furrowed, breathing fast, limbs rigid, flinching at every touch — yet the chart reads pain assessment zero. The intern asks whether the patient himself said he wasn't in pain, and the senior resident answers that he hasn't spoken in two years, so who filled in that zero he just saw. Pain by definition is subjective, an unpleasant sensory and emotional experience associated with actual or potential tissue damage, and the word potential is deliberate — it means pain can be real even with no visible damage. Since it is a subjective experience, the only gold standard is the patient's own report, but self-report carries one precondition, that the patient must be capable of reporting, and infants, patients with advanced dementia, those intubated and sedated, delirious, or severely intellectually disabled cannot clear that bar; for them the only window left is behavior — facial expression, limb tone, vocalization, consolability, and breathing pattern. So scales are not divided into accurate and inaccurate, but into those you can ask and those you can only observe, and choosing the wrong category produces a score fabricated out of thin air.

Among the self-report scales, the numeric rating scale runs from zero for no pain to ten for the worst imaginable pain, and since it can be asked verbally it is the most practical at the bedside; the visual analogue scale has the patient mark a ten-centimeter line and then measures the mark in millimeters, common in research but cumbersome at the bedside; the verbal rating scale suits older adults who are uncomfortable with abstract numbers; and there is the Wong-Baker FACES scale, suited to children roughly three and older or to those facing a language or cultural barrier. The biggest trap sits right on the faces scale — because it depicts faces, many people assume the caregiver scores the patient's expression, when in fact it is a self-report tool: the patient must point to whichever face most resembles how he feels right now, and if he cannot point, the answer is to switch scales, not to point on his behalf. Among the behavioral observation scales, FLACC is used for young children, scoring face, legs, activity, cry, and consolability at zero to two points each; PAINAD is used for those with advanced dementia who cannot speak, scoring breathing, negative vocalization, facial expression, body language, and consolability; and ICU patients who are intubated and sedated are scored with CPOT or BPS, observing the face, body movements, and synchrony with the ventilator. One more rule must never be forgotten: vital signs cannot serve as a pain indicator, since a patient with chronic pain can have a perfectly normal heart rate and blood pressure. A clinically meaningful improvement is generally taken as a drop of at least two points or at least thirty percent, and a complete pain assessment covers more than intensity alone — it must also ask about location, quality, temporal pattern, aggravating and relieving factors, and the impact on function, since recording a single number is the most common quality defect. In the mid-nineteen-nineties the American Pain Society proposed pain as the fifth vital sign, the veterans' health system adopted it in nineteen ninety-nine, and the Joint Commission implemented pain management standards in two thousand one, so that pain turned from a complaint nobody owned into a score that had to be documented, audited, and improved. What followed remains contested to this day: over that same period opioid prescribing rose sharply and evolved into a public health crisis, yet there is not enough causal evidence to pin the crisis on the fifth vital sign alone; the more defensible explanation is structural — once a score is written into accreditation standards without an equally forceful mandate on how it should be brought down, whatever lowers the score most easily gets used the most.

In the consultation room next door, the patient's daughter says please don't tell my father. Whether to disclose has already been settled earlier — the rightful recipient of disclosure is the patient himself, and a family's good intentions cannot override the patient's right to know; what remains here is the next question, how to tell it once you have decided to. In the six steps of SPIKES, setting up means arranging privacy, sitting down, silencing the pager, and confirming that everyone present is who the patient wants present; assessing perception means asking first what the patient currently understands; obtaining the invitation means asking whether he wants every detail or just the next step; giving knowledge means leading with a warning shot that the results are not what we hoped for, then delivering it in small pieces free of jargon, pausing after each to check in; responding with emotion means addressing the feeling before the information, naming the emotion, allowing silence, and expressing understanding; and only then comes strategy and summary. The reason the order cannot be swapped is cognitive load — in a moment of intense emotion a person's processing capacity drops sharply, and in the minutes after bad news is delivered the patient can barely absorb any information, so the most important content cannot come right after the emotion; the emotion must be caught first. And before delivering information you must ask about understanding and willingness, partly to know where to start, and partly because not wanting to know is itself an exercise of autonomy. The companion techniques are ask-tell-ask, and naming, understanding, respecting, supporting, and exploring. On the institutional side, Taiwan has promoted shared decision making and built a platform of decision aids since two thousand sixteen; it differs from informed consent, since informed consent is one-way disclosure followed by authorization when a recommended plan already exists, while shared decision making is two-way deliberation used when two or more reasonable options exist and their relative merit depends on the patient's values. Four communication details are worth remembering: never say there is nothing more that can be done, and instead shift the goal rather than declare defeat; silence is not a failure but a tool; how much time is left should be answered with a range and acknowledged uncertainty; and for children and adolescents, legal consent belongs to the guardian, but the child's assent should still be obtained according to his level of understanding.

Of the medical record's four columns, the most valuable is the assessment column. The subjective column records the patient's own account, including relevant negative findings; the objective column records vital signs, the physical exam, and laboratory and imaging results — both of these are data that anyone can look up afterward. The assessment column is the only place in the entire record where the reasoning process is documented, which possibilities were considered at the time, why each was excluded, why this one was chosen; the plan column records investigations, treatment, patient education, and the timing of reassessment. When a dispute arises later, what must be judged is not whether the outcome was correct but whether the judgment made at the time matched the standard of care then prevailing, and the only thing that can prove this is the reasoning written in the assessment column — so a record with a blank assessment column is, as evidence, indistinguishable from a physician who thought about nothing at all, no matter how thoroughly he actually reasoned it through at the time; carried to its conclusion, this is that same cold rule, that what is not documented is, in the eyes of the law, treated as though it was never done. Article sixty-eight of the Medical Care Act has three paragraphs worth knowing precisely: the first requires personal recording along with a signature or seal and the date it was performed; the second requires that any addition or deletion be signed or sealed at the point of change and dated, with the deleted portion struck through by a line rather than obliterated; the third requires that a verbal order be completed in writing within twenty-four hours. The second paragraph defines exactly what a lawful correction looks like — draw a line through it so the original text remains legible, write the correct content alongside it, then sign and date it; painting over an entry with correction fluid, tearing out the page and rewriting it, or backfilling the record afterward with a version more favorable to yourself turns a correction into falsification and tampering, which beyond administrative penalties may carry criminal liability as well, since what this provision protects is not tidiness but auditability. The electronic record has introduced a new risk: in the paper era, handwriting speed itself acted as a filtering mechanism, but the electronic record drove the cost of copying and templating toward zero, so each day's note became yesterday's copy with a small edit, and the result is note bloat, where length explodes while information density collapses; more dangerous still is the propagation of copied errors, since an allergy history entered wrong on day one, or an unverified assumption, gets copied intact all the way to day thirty, appearing as though it had been independently confirmed every single day, so the error accumulates a credibility it does not deserve. The remedy is discipline, not software: the assessment column must be rewritten daily, copied content verified item by item, and timestamps and authorship must never be falsified, since the audit trail leaves a trace. One more thing has changed who reads the record — the patient now reads it himself; Taiwan has offered My Health Bank since two thousand fourteen, letting the public download their own records of visits, medications, and tests, so phrasing carrying judgment, such as the patient complains or noncompliant, must change, because derogatory phrasing is absorbed by the next physician who reads the chart and goes on to shape subsequent care — it is a genuine clinical variable, not merely a matter of courtesy.

Last comes the one form everyone eventually encounters, yet almost no one is ever formally taught how to complete. The physician on call gets a phone call: an old man with end-stage liver cancer stopped breathing at three in the morning, and the pen stops over the box for the disease or injury directly causing death, when the nurse adds that the family says he fell and hit his head in the bathroom last week. A death certificate is not an administrative form; it sets in motion the handling of the body along with inheritance, insurance, and de-registration procedures, and once it is issued the body is typically cremated within a few days, so every piece of potential physical evidence disappears simultaneously and irreversibly — which is why the law sets the bar at its most conservative point. Article eleven-one of the Physicians' Act provides that a physician may not issue a death certificate or stillbirth certificate without personally examining the body, and it cannot be issued on the basis of a phone call, a family member's account, or a prior medical record; Article seventy-six of the Medical Care Act provides that a hospital or clinic may not refuse to issue a death certificate absent a statutory reason, that any diagnostic certificate must be handled with the utmost care, particularly regarding the cause of death, and that a death that is not from natural disease, or suspected not to be, must be reported to the prosecutorial authority for an inquest in accordance with the law; Article sixteen of the Physicians' Act carries the same intent. So on hearing fell and hit his head, the correct response is to put the pen down, not to keep filling out the form. An administrative inquest applies when death is confirmed as natural, is conducted by a physician who issues the death certificate, and is initiated by the family's application; a judicial inquest applies when death is not natural or suspected not to be, is conducted by a prosecutor together with a forensic physician or examiner, with an autopsy if necessary, produces an inquest certificate, and is initiated when a physician or the police report it to the prosecutorial authority. The test is not whether it looks like an accident but whether a non-natural death can be ruled out, and if it cannot be ruled out, report it for an inquest — that is the choice with the lowest threshold and the least risk. The cause-of-death section is not asking you to write down the cause of death but demanding that you write out a causal chain, where line one is the immediate cause of death, line two is the cause that produced line one, and line three is the cause that produced line two, with a separate field for other significant conditions recording diseases that contributed to death but do not sit on that chain. Because death is a process, not an event, a patient with gastric cancer may first lose the ability to swallow, then develop aspiration pneumonia, and finally die of respiratory failure, and if statistics registered only that last box, the nation's ranking of causes of death would collapse into respiratory failure, cardiac arrest, and multi-organ failure, with no public health meaning whatsoever. So the cause-of-death coding rules of the International Classification of Diseases mandate that the single code assigned is for the underlying cause of death, meaning the disease or injury sitting furthest upstream in that chain, the one without which death would not have occurred, the entry written on the bottom-most line; the underlying cause of death determines where resources flow, since cancer screening budgets, injury prevention, and chronic disease care priorities all rest on that one upstream entry, so filling it in wrong is not a slip of the pen but a turn of the public health steering wheel in the wrong direction. Cardiopulmonary failure, respiratory failure, cardiac arrest, and multi-organ failure are none of them acceptable underlying causes of death — they are modes of dying, since everyone's heart and lungs stop when they die; senility should be used only as a last resort in an elderly patient with genuinely no other identifiable cause; each line should record only one cause, and the line below must plausibly have caused the line above; an external cause must record both the mechanism and the circumstances of the injury, not merely head trauma; and the time of death must be recorded truthfully and never backdated. This chapter's throughline is a single sentence: speak it, write it, sign it — every step of it is not paperwork, but care itself.