All-cause mortality
The share of people dying from any cause over a period, regardless of cause. The book uses it to measure "how long people live". The original literature calls it all-cause mortality, abbreviated ACM
A quick reference to 41 statistical, health and legal terms used in the book. Look up a term whenever a claim needs a closer reading.
The share of people dying from any cause over a period, regardless of cause. The book uses it to measure "how long people live". The original literature calls it all-cause mortality, abbreviated ACM
Hazard ratio. It compares how quickly an event occurs at a given point during follow-up among people who have not yet experienced it. HR 0.87 means this hazard is 13% lower; it does not mean the probability of an event over a period is 13% lower.
Relative risk. The ratio of event probabilities in two groups over the same observation period. RR 0.87 means a relative reduction of 13%, not 13 fewer events in every hundred people.
Odds ratio. First calculate the probability of an event divided by the probability of no event in each group, then compare those two odds. It is not a ratio of probabilities. It approximates RR only when the outcome is rare, so it cannot always be translated into a percentage increase or decrease in probability.
Incidence rate ratio; read like RR
Rate ratio (e.g. number of falls); read like RR
SMR. Actual deaths in a group divided by the expected number computed from the mortality of the general population of the same age. 5.86 means 5.86 times the deaths of peers
The probability of an event in one group minus the other, directly giving "how many extra cases per thousand". Ratios only say how many times; the risk difference says how many more people in absolute terms
95% confidence interval. The range where the true value most likely sits. If a ratio-type interval crosses 1, or a difference-type interval crosses 0, the difference may be pure chance and the text will say "not statistically significant"
Randomized controlled trial. People are randomly split into two groups, one gets the intervention, one does not, and outcomes are compared. The best evidence of causation
Pooling the results of multiple studies into one overall estimate
Cohort study. A group of people is followed for years to see who has events. Shows association, cannot fully show causation
Observational study. Researchers only observe, do not intervene; results may be affected by confounding and reverse causation, so discount the numbers
A third factor that affects both cause and result, making association look like causation. People who eat more vegetables also tend to exercise more
Not A causing B, but B causing A. It is not that sleeping a lot makes people die early; it is that seriously ill people sleep a lot
Effect size. How many standard deviations apart two group means are. 0.2 is small, 0.5 medium, 0.8 large
Correlation coefficient. How strongly two things move together, from -1 to 1; 0.1 weak, 0.3 medium, 0.5 strong
Unit of exercise intensity. 1 MET is sitting quietly; brisk walking is about 3 to 4 MET. MET·h is intensity times hours
An international standard for grading evidence quality into high, moderate, low, and very low
Counting the effect by "who was invited" rather than "who actually went", which underestimates the effect on those who actually went
A unit of smoking amount. Packs per day times years of smoking; 30 pack-years is one pack a day for 30 years
Body mass index. Weight (kg) divided by height (m) squared
Low-density lipoprotein cholesterol, colloquially "bad cholesterol"
Estimated glomerular filtration rate. A measure of how well the kidneys filter, in mL/min/1.73 m²; the lower the number, the worse the kidney function, and chronic kidney disease is staged by it
Hepatitis B surface antigen; positive means infected with hepatitis B
Human papillomavirus; long-term infection with some types causes cervical cancer
Low-dose chest CT, with about one-fifth to one-tenth the radiation of a regular CT
Airborne particulate matter under 2.5 micrometers in diameter
A food classification by degree of processing; "ultra-processed food" is its fourth category
Loan Prime Rate. China's benchmark lending rate, published on the 20th of each month; mortgage and private-lending caps both reference it
The labor arbitration ruling takes effect directly; the employer cannot sue in court
Marriages and divorces registered in the year per thousand people. Divorces divided by marriages is another measure, the divorce-to-marriage ratio; do not mix the two
Automated external defibrillator. The red or yellow first-aid box common in public places; turn it on and follow the voice prompts, it decides by itself whether to shock
Cardiopulmonary resuscitation. Pressing hard on the chest during cardiac arrest to keep blood flowing
China Compulsory Certification. Products in the catalog may not leave the factory or be sold without this mark
The registration a website or app completes in the MIIT system before going live
China's cybersecurity classified protection system; protections are applied according to how important the system is
An open-source license. Build a product with its code and you generally must open-source your product when distributing
An agreement not to join a competitor for a period after leaving; the company must pay monthly compensation, capped at 2 years
The money promised when registering a company. Once promised it must actually be paid in within the statutory deadline; it is not just for show
A statutory deposit (dingjin) carries a penalty: the holder who breaches returns double, capped at 20% of the contract; a prepayment (yudingjin) carries no penalty
The book refers to the legal time limit for the court, not the actual waiting time. In an ordinary first-instance civil case, the court must finish within six months after docketing; an extension may be approved in special circumstances. The summary procedure has a three-month limit. Time spent on expert appraisal and service by public notice is excluded, so the actual duration is often longer.