Increasing comorbidity is associated with worsening physical function and pain after primary total knee arthroplasty
TKA is the most common arthroplasty procedure worldwide and its utilization is increasing rapidly [1, 2]. Therefore, it is important to understand the variability in TKA outcomes. We developed and tested a novel medical/surgical comorbidity index, the Minnesota Arthroplasty Comorbidity Index (MACI). MACI took into account a greater number of comorbidities as compared to the Charlson index, weighted severity of conditions not only presence (medical index subscale) and included local musculoskeletal index and TKA-related index subscales. We examined the longitudinal association of medical comorbidity, local musculoskeletal comorbidity and index TKA-related morbidity with worsening of pain and function post-TKA.
As expected and has been shown previously, pain and function scores improved dramatically from pre-operative to the post-operative interval. Most, but not all these gains persisted during intermediate-long post-TKA mean follow-up of 4.9 years post-operatively (range, 1.3–11.4 years). We found that medical and local musculoskeletal comorbidity in the post-TKA period were associated with gradual worsening of function and pain outcomes post-TKA, after a striking initial improvement with TKA, as demonstrated previously with observational studies. Several findings merit further discussion.
A few previous studies have reported that pre-operative comorbidity predicts short-term or very short-term post-arthroplasty outcomes. SooHoo et al. found that the preoperative Charlson score predicted several short-term measures including 90-day readmission postoperatively [12]. Weaver et al. found that in 11,710 veterans with TKA, higher preoperative Deyo-Charlson score (modified version) correlated with longer length of stay and complications 30-days postoperatively [30]. In a study of relatively healthy Australian patients undergoing total hip or knee arthroplasty, baseline individual comorbidity was predictive of change in SF36 physical domain score from the pre- to post-operative period 12-months post-arthroplasty [31]. A previous study showed that SF36 and WOMAC scores declined gradually over the next several years in a study of 551 TKA patients, after an initial post-operative improvement [7]. This decline correlated with pre-operative baseline comorbidity [32]. What we don’t know, however, is whether an increase in comorbidity through time correlates with the decline in SF36 and WOMAC scores post-TKA. Our study advances knowledge by examining the association of change in comorbidity longitudinally after primary TKA long-term with pain and function outcomes post-TKA. In our study, we examined change in comorbidity post-operatively over time, and examined that as the predictor of pain and function in subsequent periods of observation. We found that increasing comorbidity post-TKA was significantly associated with future worsening of pain and physical function after primary TKA.
Our findings must be compared to a cross-sectional study done previously on this topic. Dunbar et al. performed a cross-sectional survey 6.7 years post-TKA patients, and used the modified patient-reported Charnley Classification [33]. The Class C patients (unilateral TKA and remote arthritis and/or a medical condition that affected their ability to ambulate) had statistically significantly worse QOL scores demonstrating that a simplified measure of comorbidity correlated with a decline in patients’ QOL. The cross-sectional study could not address this relationship beyond a correlation, since there was only one measurement post-TKA and both comorbidity and QOL were examined at the same time. Our study extends these findings of association of post-TKA comorbidity with post-TKA outcomes, by studying them longitudinally for the first time, controlling for baseline post-operative pain/function and other confounders, assessing the worsening or comorbidity in periods prior to the assessment of pain and function and using multiple measures of comorbidity. These findings are important for several reasons.
The clinical relevance of our study findings is that increasing comorbidity in post-TKA years might explain declining pain and function outcomes in patient subgroups ?in the years after TKA. Surgeons and patients should discuss this during the informed consent process, so that expectations are realistic and patients are empowered with this information. Close monitoring and early intervention to treat new/worsening comorbidities might help in reducing the impact of comorbidities on pain and function outcomes in the long-term follow-up post-TKA.
Instead of a more simplified measure as described above, we used two weighted measures of comorbidity in our study, the Charlson index, and our novel arthroplasty comorbidity severity index. The method of measuring and reporting comorbidities in previous studies varies significantly. Many comorbidity indices tabulate the comorbidities as a single score with a sum. Studies commonly stratify patient comorbidity sum as dichotomous variable, e.g. 4 vs. 4 or more comorbidities [34]. The summation and stratification methods, however, do not account for the severity of various illnesses. To address this issue, weighted, validated comorbidity measures have been created, the best known of which is the Charlson index [22, 35]. Charlson index correlates with mortality, hospital readmission, length of stay, post-operative complications, progression-free survival in cancer patients, and disability [35, 36]. It has been validated in various conditions, e.g., pneumonia, heart disease, spine surgery and amputation [36].
In this study, we explored the relationship between comorbidity and postoperative worsening of QOL measures. We examined a more comprehensive spectrum of comorbidities than the Charlson index by creating a novel medical/surgical comorbidity index focusing on common comorbidities of TKA patients, which consisted of three component subscales: medical, local musculoskeletal (lower extremity and spine morbidity), and TKA-related comorbidities (e.g. problems with the TKA itself). Our medical/surgical index “medical” subscale examined 25 common comorbidity categories compared to 13 in the Charlson index, and while there is some overlap, we included several more common diagnoses in the ambulatory patient populations such as those undergoing TKA (e.g. hypertension), rather than more severe (but much less common) diagnoses as in the Charlson index (e.g. AIDS or “second metastatic solid tumor”); the score ranges 0-47. We also included several diagnoses of a more function-limiting nature, such as hearing and vision problems or alcohol abuse. We found that, both the Charlson Index and our “medical” subscale score from the novel index correlated with worsening SF36 PF and WOMAC PF. Future studies should compare their discriminative abilities for TKA outcomes to understand how to best use these measures in arthroplasty patient populations.
Interestingly, our local musculoskeletal index subscale score also correlated significantly with SF36 PF and WOMAC PF. This index included opposite knee, hip, and spine arthritis, as well as local problems such as neuropathy and vascular disease. This suggests that lower extremity and spine issues, though localized, are important determinants of lower extremity as well as global physical function. Not surprisingly, studies have shown that ipsilateral hip arthritis/pain, low back pain, and contralateral knee arthritis/pain might contribute to index knee pain [37–40]. And because the non-overlapping TKA-related Index included all TKA-related problems such as loosening or infection, this finding the association of local musculoskeletal index subscale score with SF36 PF and WOMAC PF cannot be attributed to the arthroplasty itself.
We also found that patients with worsening local musculoskeletal index subscale score had more pain in the subsequent periods post-operatively, indicated by a significant association with SF-36 pain scores. Pain correlates with worse functional outcome [41]. Post-TKA pain is likely multifactorial, related to more than just the index knee replacement. Increasing comorbidity post-TKA may interfere with optimal rehabilitation leading to suboptimal recovery and residual pain and functional limitation. It remains to be seen whether targeted pain control and rehabilitation protocols, and optimal perioperative management of comorbid conditions can impact post-TKA pain and function outcomes.
Our study has some limitations. We had a relatively small sample size of 124 patients, of which 123 were men. We considered limiting the analyses to only men, but decided to analyze the entire sample, since it makes our findings more representative of the entire VA patient population (95–99 % men and 1–5 % women), not just male veterans. However, we conducted sensitivity analyses limited to men, which had the same conclusions as the total cohort (as expected). We controlled for age, length of time since surgery, and baseline QOL values, but not for other factors. Some patients, for example, had undergone only unilateral TKA, while others had (either before or after the index surgery) undergone second TKAs on the opposite knee. This could have impacted the pain and function scores, particularly in patients who had contralateral arthritis but had not had a contralateral TKA; it would be interesting to explore this relationship. We did, however, attempt to control for this by measuring lower-extremity-related comorbidity as described above. Our patients had been enrolled in three different studies, two randomized and one cohort, which may have introduced sample heterogeneity and we were not able to control for it. The mean follow-up interval prior to each assessment was 12–20 months; while slight differences in time-periods can lead to some heterogeneity in parameter estimates, we used a repeated measures mixed model linear regression and comorbidity assessments were performed for the period prior to each QOL assessment, partially accounting for this.
Comorbidities were abstracted by a physician’s chart review of their documentation in health care provider notes. The accuracy of this approach of comorbidity assessment in a retrospective study can not be assessed due to the absence of a gold standard, such as patient examination in a prospective cohort study. Despite suspected under-documentation of some comorbidities (especially psychological), we suspect that this method of capturing comorbidity may be more accurate than capturing diagnostic codes for comorbidity from administrative databases. Lastly, our newly developed scale is not validated as yet; test of reliability and several other aspects of validity are needed before any further use of the MACI scale. However, additional value over the existing comorbidity scales wasn’t evident in this study cohort. It is possible that this scale may be more relevant for patients undergoing revision arthroplasty or other joint procedures.
Our study has several strengths. First, our population was well suited to this investigation. Medical comorbidities are prevalent in the veteran population, since 72 % of VA patients have at least one chronic condition [37]. This is significantly higher than the 47 % of Americans overall with at least one chronic condition [38]. We performed a complete review of both inpatient and outpatient records. Zhang et al. reported that compared to a look-back period of 1 year of inpatient data, a longer look-back period using multiple sources of data (2 years of inpatient plus 1 year of outpatient) significantly improved the mortality prediction of the Charlson index [39]. We reviewed more data than this in our patients, with an average look-back period of 10.5 years (range 2.8–33.8, SD 3.26 years), examined a long-period for assessing baseline comorbidity. This is a longer follow-up than in many post-arthroplasty studies of comorbidity, which mostly ranged from 30 days to 2 years [9, 30, 34, 40, 42, 43]. In comparison to certain physician-derived scores such as the Knee Society Score, which requires a surgeon’s assessment, [44] we used validated, patient self-reported instruments (the WOMAC and SF36) to measure post-TKA outcomes. In TKA patients, WOMAC may have a bigger effect size and overall responsiveness than generic surveys [45] such as SF-36 and better correlate with knee-specific complaints [46].