Berlin. The global economy could generate an additional $1 trillion annually if the medical data gap between women and men could be closed. The McKinsey Health Institute has calculated this for the year 2040. A new white paper, which is available exclusively to Handelsblatt, takes up the estimate and shows specifically how male-biased research data worsens medical care for women - and that this effect is further intensified by the use of artificial intelligence (AI).
When making diagnoses and treatments, doctors still rely on findings that researchers have mainly gained from men. As a result, they may recognize certain diseases in women later or prescribe medications whose effect on the female body has not been sufficiently studied.
The data gap runs through all medical research, write biotechnologist Nina Haffer and economics professor Anabel Ternès-von Hattburg in their white paper “We Highlight what Systemic Bias Misses”. Pharmaceutical companies, medical technology manufacturers and hospitals should not only create equal treatment for moral reasons. The paper states that you must also understand insufficient health data as a business risk and systematically improve its quality.
“Better data is an important basis for making health inequalities visible and quantifying them more precisely,” says Reinhard Busse, professor of health care management at the Technical University of Berlin and health economist. “Especially if we want to use routine data, the crucial characteristics must also be recorded.”
Measures could then be derived to eliminate the inequalities. “Only then can we estimate what health and economic benefits we can achieve as a result,” says Busse.
The causes of the insufficient data sometimes go back decades. For a long time, scientists predominantly used male laboratory animals in drug research. The white paper cites research that shows studies with all-male rodents were about 5.5 times more common than those with all-females.
Researchers also sometimes excluded women from clinical studies, partly because of concerns about the effects of experimental drugs on pregnancies. Nevertheless, findings from tests with men were incorporated into reference values and treatment recommendations that later applied to both sexes.
In fact, hormones or metabolism influence how the body absorbs and breaks down medications. If pharmaceutical companies do not adequately investigate these differences, they only discover gender-specific side effects after approval. They then have to add additional warnings to their medication or take it off the market.
The symptoms of a heart attack show particularly clearly how the data gap affects medical care. When a coronary artery is blocked, women have different accompanying symptoms than men, such as nausea or unusual exhaustion. Certain laboratory values also differ. If gender-specific limits are set in the blood test for the heart attack marker troponin, the detection of a heart attack in women could be improved, says the white paper with reference to studies.
Artificial intelligence threatens to exacerbate these problems. Hospitals and medical technology companies are increasingly relying on AI to detect diseases earlier and reduce the burden on doctors. But algorithms learn from existing data. If these predominantly depict male patients, the systems may be less able to detect diseases in women.
Sylvia Thun, data scientist at the Berlin Institute of Health, warns: “AI can deliver very good results overall and still overlook women more often.” That's why we should no longer just look at the average hit rate in the future. “We have to examine the error rates separately for women and men, otherwise there is a risk that AI will not eliminate existing biases in medicine, but will automate it,” she says. Thun was involved as an author on the white paper.
The authors also identify weak points in the digital recording of medical information. They warn that international classification systems sometimes lack suitable categories or that diseases are assigned inconsistently. Such gaps make statistical analysis difficult and can impact the development of medical AI.
This is likely to increase political pressure to eliminate these risks. The European AI Act already places high demands on certain medical AI systems, such as data quality and risk management. However, the timing of individual requirements continues to be discussed.
