There’s one big problem with follow-up studies in medicine: patients who are “lost to follow-up.” These patients are included in the study at the outset. But then they stop returning letters or phone calls, or just generally make themselves scarce. These “lost” patients can sway the results of studies if, as a group, their outcomes are not like the patients who stayed in the study. Luckily, most studies lose few patients to follow-up.
However, patients who don’t respond are more of a problem in mail surveys. And mail surveys are being used more and more by hospitals and doctors. After total knee replacement (TKR), mail surveys are often used to rate patient satisfaction and knee function. In this study, mail surveys were sent to 472 TKR patients. The survey included 10 simple questions. The answers were compared to the doctors’ records from the patients’ clinic visits. But the focus of the study was not so much on the answers. These researchers wanted to know if the nonresponders would in any way have changed the outcomes.
To do this, the researchers needed to make sure that all 472 patients answered the survey questions. For 83 percent of the patients, this was not a big problem. They returned their surveys after one or two mailings. The remaining 17 percent took some nagging. These 80 nonresponders got up to four further mailings and then phone calls until they answered the questions.
When the results were all in, the data was sorted by how quickly patients had responded. There were important differences between early responders and nonresponders. Early responders generally had:
In a normal study, the nonresponders would not have been hounded until they answered the survey. In this case, that means the survey would have shown much more positive results than were actually true. The patients “lost to follow-up” would definitely have swayed the survey results to make the outcomes of TKR look better.
The study also showed that the clinical scores recorded by doctors during clinic visits looked better than patients’ reports. The authors recommend that studies should allow patients to report their own levels of subjective items such as pain. The authors also recommend that all surveys should make every effort to get all patients to answer surveys. It could make a big difference in the final data.