Research glossary

The proportion of people in a group who die during a given period, whatever the cause. This handbook uses it when discussing longevity. Often abbreviated ACM.

Hazard ratio: a comparison of how quickly events such as illness or death occur in two groups over time. An HR of 0.87 indicates a 13% lower hazard; 1.21 indicates a 21% higher hazard.

Relative risk: the probability of an event in one group divided by that in another. Read the ratio in the same general way as an HR.

Odds ratio: another way to compare two groups, using odds rather than probabilities. For rare outcomes it is close to the RR; for common outcomes, treating it as an RR exaggerates the difference.

Incidence rate ratio: the rate of new events in one group divided by the rate in another. Ratios below or above 1 indicate lower or higher rates, as with an RR.

Rate ratio: compares the number of events, such as falls, rather than simply how many people experience an event. Read the ratio in the same general way as an RR.

SMR: observed deaths divided by the deaths expected in a comparable general population after standardization, such as for age. An SMR of 5.86 means 5.86 times the expected number.

One group’s event probability minus the other’s. It can be expressed as the actual number of additional cases per 1,000 people, rather than a relative multiple.

A 95% confidence interval expresses uncertainty around an estimate. If an interval for a ratio includes 1, or an interval for a difference includes 0, the data are compatible with no difference; this is commonly described as not statistically significant.

Randomized controlled trial: participants are assigned by chance to intervention and comparison groups, then their outcomes are compared. Random assignment provides particularly strong support for assessing whether an intervention caused a change.

A statistical combination of results from multiple studies to produce an overall estimate.

A study that follows a group of people, often for years, and records outcomes. It can identify associations, but an association alone does not establish causation.

Researchers observe rather than assign the exposure or treatment. Other factors can distort the findings; see confounding and reverse causation. Interpret such estimates with that limitation in mind.

A third factor affects both the exposure and outcome, making a relationship look causal. People who eat more vegetables may also exercise more, making it difficult to separate the effects.

The apparent direction of cause and effect is reversed. For example, illness may make people sleep longer, rather than longer sleep causing their higher mortality.

Standardized effect sizes, expressing the difference between groups in units of typical variation. Rough conventions are 0.2 for a small effect, 0.5 for medium and 0.8 for large.

Correlation coefficient: how closely two quantities vary together, from −1 to 1. Common rough benchmarks for magnitude are 0.1 small, 0.3 moderate and 0.5 strong.

A measure of activity intensity. Sitting quietly is about 1 MET; brisk walking is roughly 3–4 METs. MET·h multiplies intensity by hours to express the amount of activity.

An internationally used approach to assessing certainty of evidence, with high, moderate, low and very low categories.

An analysis based on who was invited to screening, not just who attended. Because some invitees do not attend, its estimate can be smaller than the benefit among those actually screened.

A measure of smoking exposure: packs per day multiplied by years smoked. One pack daily for 30 years is 30 pack-years.

Body mass index: weight in kilograms divided by height in metres squared. It is commonly used as an indicator of body size.

Low-density lipoprotein cholesterol, commonly called “bad” cholesterol.

Estimated glomerular filtration rate, a measure of kidney filtration in mL/min/1.73 m². Lower values indicate poorer filtration and help determine chronic kidney disease stage.

Hepatitis B surface antigen, a laboratory marker. A positive result indicates hepatitis B virus infection.

Human papillomavirus. There are many types; persistent infection with some types can lead to cervical cancer.

Low-dose chest computed tomography, with radiation exposure roughly one-fifth to one-tenth that of a conventional CT in the comparison used here.

Fine particles suspended in the air with a diameter of 2.5 micrometres or less.

A food classification system with four groups based on the nature and extent of processing. Ultra-processed foods are group four.

China’s loan prime rate, a lending benchmark announced monthly on the 20th. It is used in mortgage pricing and in calculating the legal interest-rate ceiling for relevant private loans in China.

A category of Chinese labor-arbitration award that takes effect without the employer being entitled to bring an ordinary new court action over the same dispute.

The number of couples registering marriage or divorce in a year per 1,000 people. The divorce-to-marriage ratio instead divides divorces by marriages in that year; it is a different measure.

Automated external defibrillator. Often housed in a red or yellow emergency cabinet in public places. Switch it on and follow the prompts; the device determines whether a shock is appropriate.

Cardiopulmonary resuscitation. Chest compressions help keep blood moving during cardiac arrest.

China Compulsory Certification, also called 3C. Products covered by the scheme must meet its certification and marking requirements before leaving the factory or being sold.

A registration procedure in China’s Ministry of Industry and Information Technology system for websites or apps before they go online, where the filing rules apply.

China’s graded cybersecurity protection system. Systems are assigned levels according to their importance, with more demanding safeguards at higher levels.

The GNU General Public License, a software license that generally requires covered distributed works to make their source code available under its terms.

In the Chinese employment context discussed here, an agreement restricting work for competitors for a period after departure. The employer pays monthly compensation during the restriction, which may last no more than two years.

Money a shareholder commits to contribute when registering a company. It must actually be paid within the legal deadline; it is not merely a number on a form.

The Chinese terms dingjin (定金) and dingjin (订金) sound alike but have different effects. The former carries a deposit penalty: a recipient who defaults may owe double repayment, and the deposit is capped at 20% of the contract price. The latter is simply an advance payment without that deposit penalty.