Patients who have total knee replacement are advised to have annual or
biannual follow-up evaluations with radiographs. Follow-up is important
because early identification of problems, such as polyethylene wear, can be
addressed early and hopefully can obviate the need for more extensive surgery.
Follow-up is also important for outcome studies. Debate is ongoing regarding
the statistical processing of patients who are lost to
follow-up1-13.
Some have argued that patients who are lost to follow-up may be more likely to
have a worse outcome and to have had surgical intervention
elsewhere7-9,13.
If this is the case, patients lost to follow-up should be considered as having
had a failure in the statistical analyses of outcome studies. However, the
assumption that patients who were lost to follow-up had a poor outcome may
incorrectly bias such studies in a negative fashion.
Using patient records and Internet search techniques, we attempted to
locate all patients, in a consecutive series of individuals treated with total
knee replacement by one surgeon, who met our definition of not having returned
for follow-up. We then compared the functional outcomes of those patients with
the outcomes of patients who did return for follow-up. Our hypothesis was that
patients who did not attend prescribed follow-up appointments were not more
likely to have undergone revision surgery or to have a worse outcome compared
with patients who did attend follow-up appointments. Furthermore, we
hypothesized that Internet search techniques would be a useful tool for
locating patients who had not returned for follow-up appointments.
After approval by the institutional review board, we retrospectively
reviewed the records of 161 patients with a total of 200 consecutive total
knee replacements performed between April 1996 and July 1997 by the same
surgeon (R.D.S.) with the same prosthesis (PFC Sigma; DePuy, Warsaw,
Indiana).
The minimum duration of follow-up was five years (mean, 64.0 months; range,
sixty to seventy-three months). All of the operations were performed at one of
two hospitals, and all follow-up examinations were conducted at the same
office. For the purposes of this study, we defined patients as not having
returned for follow-up if they had had no contact of any type with their
surgeon beyond six months after the date of the surgery.
An attempt was made to locate patients who had not returned for follow-up
at a minimum of five years by using their last known contact information or
the last known information on their next of kin. When a patient could not be
located with use of this information, a series of searches of free, readily
available Internet databases was carried out with use of the patient's most
recent demographic information as a starting point
(Table I). A standardized
algorithm was used for all patients (Fig.
1)14.
Patients who had not returned for follow-up were evaluated by one of the
authors (A.S.M.), who had not been involved in their care. The evaluation was
carried out by means of a telephone interview, during which the patients were
asked about the status of the knee prosthesis, the reason that they did not
return for follow-up, and whether a different surgeon had been evaluating or
treating the knee. A patient who gave more than one reason for not adhering to
the recommended follow-up regimen was asked to identify which reason he or she
considered to be primary. Scores for the pain and function components of the
Knee Society Clinical Rating System were determined on the basis of this
telephone interview and were compared with the preoperative values recorded in
the patient's
chart15. Patients
were educated about the importance of the recommended follow-up regimen. All
patients were asked to schedule an appointment for complete physical
examination and radiographs.
Patients who had been returning for follow-up appointments were evaluated
in the same fashion. If the patient had already returned for a follow-up
appointment at a minimum of five years, the Knee Society pain and function
scores were determined from the chart. If the patient had been keeping
follow-up appointments but had not yet returned for the five-year evaluation,
he or she was contacted by one of the authors (A.S.M.), who administered the
pain and function components of the Knee Society Clinical Rating System in a
telephone interview. The scores were compared with the preoperative values
recorded in the chart. All patients who had not yet returned for a clinical
evaluation at a minimum of five years were asked to return for clinical and
radiograph examination. When a patient had been keeping the prescribed
follow-up appointments but had not yet returned for the five-year follow-up
evaluation and could not be contacted with use of the last known contact
information in the chart, an attempt was made to locate that patient with use
of the standardized Internet search algorithm employed for the patients who
had not returned for follow-up.
Six knees in six patients who had returned for follow-up at a minimum of
five years had required additional surgery. The Knee Society scores for these
six knees were excluded from the analysis.
Statistical analysis was performed with use of the Student t test and the
Pearson chi-square test as appropriate, and significance was considered to be
a p value of <0.05. Power calculations indicated that a sample size of
twenty-three patients designated as not returning for follow-up provided 90%
power for determining differences in mean age, weight, and preoperative and
postoperative pain and function scores between the two groups with use of the
unpaired two-tailed Student t test. Power calculations were performed with use
of nQuery Advisor (version 4; Statistical Solutions, Boston, Massachusetts)
(a = 0.05, ß = 0.10, effect size = 1.0).
Thirty patients with thirty-five replaced knees met the definition of not
having returned for follow-up. Their mean age at the time of the surgery was
71.3 years (range, forty-four to eighty-three years), and their mean weight
was 82.0 kg (range, 52 to 111 kg). Nine (30%) of the patients were male. The
preoperative diagnoses included osteoarthritis (twenty-seven patients),
rheumatoid arthritis (one), osteonecrosis (one), and osteoarthritis secondary
to septic arthritis (one). Preoperatively, twenty-two knees had a varus
deformity and thirteen had a valgus deformity. Seven patients (eight knees)
had died of unrelated causes, at a mean of 32.4 months (range, eight to
fifty-four months) after the total knee replacement, leaving twenty-three
patients with twenty-seven knees available for follow-up. All twenty-three
patients were located, and all knees were evaluated at a minimum of five years
(mean, 63.9 months; range, sixty to sixty-nine months).
One hundred and thirty-one patients with 165 treated knees who did return
for follow-up had a mean age at the time of surgery of 68.1 years (range,
forty to eighty-four years) and a mean weight of 79.0 kg (range, 30 to 130
kg). Forty-five (34%) were male. The preoperative diagnoses included
osteoarthritis (111 patients), rheumatoid arthritis (twelve), osteonecrosis
(one), posttraumatic osteoarthritis (four), psoriatic arthritis (two), and
spondyloepiphyseal dysplasia (one). Preoperatively, 121 knees had a varus
deformity and forty-four had a valgus deformity. Ten patients with eleven
treated knees had died of unrelated causes, at a mean of 42.4 months (range,
twenty-two to sixty-four months) following surgery, leaving 121 patients with
154 knees available for follow-up at a minimum of five years. Two patients
(two knees) did not complete the questionnaire for the Knee Society Clinical
Rating System but did report that they were asymptomatic and had not had
revision surgery at a minimum of five years postoperatively. One patient (two
knees) who had been keeping appointments could not be located for the
evaluation at a minimum of five years postoperatively.
There was no significant difference between the patients (knees) who had
not attended follow-up evaluations and those who had with regard to sex,
diagnosis, deformity (varus or valgus), or weight. The patients who had not
attended follow-up appointments tended to be older at the time of the surgery
than those who had attended follow-up appointments (mean age, 71.3 compared
with 68.1 years, p = 0.10) and more of them had died (p = 0.01).
Of the thirty patients who had not returned for follow-up, thirteen (43%)
with sixteen replaced knees could not be located with use of information
contained in the chart, but all were located with a variety of Internet
searches performed with an established search algorithm. Four of these
patients (five knees) had died. The remaining patients were contacted.
Twenty-six (20%) of the 131 patients who had been keeping their prescribed
follow-up appointments had not yet returned for their follow-up evaluation at
a minimum of five years and could not be located with use of their last known
contact information contained in the chart. Twenty-four of these patients were
located with use of the Internet search algorithm. Eight of them had died.
Overall, thirty-seven (95%) of the thirty-nine patients who could not be
located with use of the last known contact and demographic information
contained in their chart were found with the Internet search algorithm. One of
the two patients who could not be located subsequently returned for follow-up
without having been contacted.
None of the patients who had not returned for follow-up had required
additional surgery on the knee; six patients who had returned for a follow-up
evaluation at a minimum of five years had required additional surgery on the
knee. This difference was not significant. The reasons for additional surgery
included late infection (two patients), arthroscopic manipulation (two),
synovectomy with exchange of the polyethylene liner because of recurrent varus
deformity (one), and excision of a lateral joint line cyst (one).
The reasons that the patients gave for not returning for follow-up are
listed in Table II. Both the
patients who had returned for follow-up and those who had not had a
significant improvement in the postoperative scores for the pain and function
components of the Knee Society Clinical Rating System compared with the
preoperative values (p < 0.0001). There was no significant difference in
the pain and function scores at a minimum of five years between the patients
who had and those who had not attended follow-up appointments.
Two patients who had not returned for follow-up appointments and four who
had returned reported that they were dissatisfied with the knee replacement (p
= 0.25). Of the two dissatisfied patients who had not returned for follow-up
appointments, one had bilateral osteoarthritis of the knee and complained of a
limb-length discrepancy following correction of a large varus deformity in one
of the knees. The other patient complained of residual stiffness one month
after the surgery and did not return for follow-up again.
There was no difference between patients who had and those who had not
attended follow-up appointments in this consecutive series of total knee
replacements evaluated at a minimum of five years postoperatively. Knee
Society scores for pain and function were similar, and no patient who had not
attended follow-up appointments had required revision surgery.
Loss of patients to follow-up is a common and unavoidable problem in
outcome studies. Much has been written about the statistical handling of such
patients1-5.
The problem is intensified when long-term follow-up is required, particularly
when the prevalence of failure is low, as is the case with total joint
replacement. If patients who are not available for follow-up evaluation are
expected to have a worse outcome, as some have suggested, authors of
orthopaedic outcome studies who do not account for these patients may report
results that are more favorable than should be
expected7-9,11,12.
Joshi et al. recently reported on a series of 563 consecutive knee
replacements and found no significant difference between the outcomes of
patients who had and had not attended follow-up
visits16. Konig et
al. found, in a series of knee replacements in Germany, that patients who had
not returned for follow-up tended to be older at the time of surgery and to
have more medical problems at the time of follow-up but none had required
revision surgery10.
Dorey and Amstutz reported similar results in a series of patients with total
hip arthroplasty6.
Others have refuted these findings but have relied on data from studies
investigating neuroses and with patients with heart
disease7. Murray et
al. reported on a series of patients treated with hip replacement and
concluded that those who were lost to follow-up could be expected to have
poorer outcomes9.
However, those authors compared data in a matched control group of patients
who had returned for follow-up with information derived from the last known
evaluation of those who had not. The current study and the study by Joshi et
al. both involved patients treated by a single surgeon, and both we and Joshi
et al. actively sought out patients who had not attended follow-up visits. Of
all of the patients in the two studies, which included 763 total knee
replacements, only one patient with two knee replacements was not located.
These comprehensive follow-up studies, unlike the study by Murray et al., do
not require any assumptions regarding the outcome of patients who were not
located. Although Dorey and Amstutz used a different methodology in their
study on total hip arthroplasty, their conclusion was the same as ours.
We agree with Laupacis that complete follow-up is certainly preferable when
possible7.
Appropriate follow-up may help to avoid complications associated with a delay
in the diagnosis of wear or loosening of a total knee
replacement17.
Unfortunately, in studies of total joint arthroplasty with long follow-up
intervals, locating patients can be difficult, time-consuming, and cost
prohibitive. Joshi et al. used a full-time research assistant and a private
detective in order to obtain follow-up data on all of the patients in their
series16. Our study
is unique in that "lost" patients were located without substantial
difficulty, time expenditure, cost, or utilization of private detectives or
other outside resources. Patients who could not be located with use of the
contact information in their chart were found with simple, readily available,
and free Internet search engines and use of a standardized search algorithm.
We therefore proved our second hypothesis that the Internet is a useful tool
for locating patients. The technique was time efficient and successfully
located 95% of the patients who had not returned for follow-up and who could
not be located with use of the last known contact information.
The wide availability and ease of use of the Internet make this modality
attractive for researchers, and its efficacy should improve as the search
engines become more sophisticated and easier to use. Search engines are easy
to use but vary in their
efficacy14. Using
an Internet search engine that referenced the Social Security Death Index
allowed us to determine which missing patients had died, and this saved
substantial time and resources by immediately ending additional attempts to
locate those patients. The proliferation of information on the Internet should
continue to make it easier to find patients. However, important issues
regarding patient confidentiality and privacy need to be considered as the use
of the Internet expands in the health-care setting.
On the basis of our results, we believe that patients who do not attend
follow-up appointments in studies of total joint arthroplasty should not be
assumed to have a worse outcome than those who do attend such appointments.
Follow-up rates can be improved by evaluating the reasons that patients do not
return and by using the Internet to locate patients. The Internet was
extremely effective in locating patients for this study, and it may prove
useful for finding patients "lost to follow-up" in a wide variety
of disciplines.