Forecasting is a construct in which experiences and beliefs inform a projection of future outcomes. Current efforts to predict postoperative patient-reported outcome measures such as risk-stratifying models, focus on studying patient, surgeon, or facility variables without considering the mindset of the patient. There is no evidence assessing the association of a patient’s forecasted postoperative disability with realized postoperative disability. Patient-forecasted disability could potentially be used as a tool to predict postoperative disability.
We completed a prospective, longitudinal study to assess the association between forecasted and realized postoperative pain and disability as a predictive tool. One hundred eighteen patients of one hand/upper extremity surgeon were recruited from November 2016 to February 2018. Inclusion criteria for the study were patients undergoing hand or upper extremity surgery, older than 18 years of age, and English fluency and literacy. We enrolled 118 patients; 32 patients (27%) dropped out as a result of incomplete postoperative questionnaires. The total number of patients eligible was not tracked. Eighty-six patients completed the preoperative and postoperative questionnaires. Exclusion criteria included patients unable to give informed consent, children, patients with dementia, and nonEnglish speakers. Before surgery, patients completed a questionnaire that asked them to forecast their upper extremity disability (DASH [the shortened Disabilities of the Arm, Shoulder and Hand] [QuickDASH]) and pain VAS (pain from 0 to 10) for 2 weeks after their procedure. The questionnaire also queried the following psychologic factors as explanatory variables, in addition to other demographic and socioeconomic variables: the General Self Efficacy Scale, the Pain Catastrophizing Scale, and the Patient Health Questionnaire Depression Scale. At the 2-week followup appointment, patients completed the QuickDASH and pain VAS to assess their realized disability and pain scores. Bivariate analysis was used to test the association of forecasted and realized disability and pain reporting Pearson correlation coefficients. Unpaired t-tests were performed to test the association of demographic variables (for example, men vs women) and the association of forecasted and realized disability and pain levels. One-way analysis of variance was used for variables with multiple groups (for example, annual salary and ethnicity). All p values < 0.05 were considered statistically significant.
Forecasted postoperative disability was moderately correlated with realized postoperative disability (r = 0.59; p < 0.001). Forecasted pain was weakly correlated with realized postoperative pain (r = 0.28; p = 0.011). A total of 47% of patients (n = 40) were able to predict their disability score within the MCID of their realized disability score. Symptoms of depression also correlated with increased realized postoperative disability (r = 0.37; p < 0.001) and increased realized postoperative pain (r = 0.42; p < 0.001). Catastrophic thinking was correlated with increased realized postoperative pain (r = 0.31; p = 0.004). Patients with symptoms of depression realized greater pain postoperatively than what they forecasted preoperatively (r = -0.24; p = 0.028), but there was no association between symptoms of depression and patients’ ability to forecast disability (r = 0.2; p = 0.058). Patient age was associated with a patient’s ability to forecast disability (r = .27; p = 0.011). Catastrophic thinking, self-efficacy, and number of prior surgical procedures were not associated with a patient’s ability to forecast their postoperative disability or pain.
Patient-reported outcome measures (PROMs) after surgery are influenced by numerous patient factors apart from clinical treatment. For example, higher socioeconomic status, better psychologic well-being, and higher patient optimism are associated with better patient-reported outcomes and quality of life [18, 23, 33, 34]. As such, understanding factors that influence postoperative PROMs can inform interventions directed toward improving outcomes, such as disability. For example, smoking cessation programs can be implemented to mitigate the effects of nicotine on nonunion rates after fracture fixation to improve PROMs . Moreover, understanding patient factors that influence postoperative PROMs, such as depression, may be useful for risk stratification modeling for PROM-based performance measures . No instrument, however, currently captures the effects of patient expectations along with their underlying psychosocial determinants of health on PROMs.
Patients may be better at predicting illness than their physician based on their own self-rated health . For example, patient self-rated health has been shown to be predictive of adverse events in patients receiving cardiac resynchronization therapy and mortality in patients with chronic heart failure, as well as outcomes in patients with type 2 diabetes and myocardial infarction [19, 21, 38].
With increasing evidence demonstrating that patients’ understand of their own health may be predictive of outcomes, they may be able to forecast their outcomes after surgical interventions. Forecasting is a construct in which experiences and beliefs inform a projection of future outcomes [2, 24]. Forecasting has previously been used to predict pain after foot surgery, as well as pain and symptoms in mastectomy patients; however, it has not been used to predict patient reported disability or function [22, 26, 27]. If patients are able forecast their postoperative disability, it may prove to be a valuable preoperative predictive tool for physicians and health systems to identify patients at risk for experiencing greater disability than expected. However, a patient’s ability to forecast their disability after orthopaedic surgery has not been studied.
We completed a prospective, longitudinal study at a suburban, academic medical center after obtaining institutional review board approval. Patients of one hand/upper extremity surgeon were recruited from November 2016 to February 2018. We enrolled participants meeting the inclusion criteria until our required sample size was achieved, accounting for dropouts. Inclusion criteria for the study were patients undergoing hand or upper extremity surgery, older than 18 years of age, and English fluency and literacy. We enrolled 118 patients and had 32 patients (27%) drop out as a result of incomplete postoperative questionnaires. The total number eligible patients was not tracked. Eighty-six patients completed the preoperative and postoperative questionnaires. Exclusion criteria included patients unable to give informed consent, children, patients with dementia, and nonEnglish speakers.
Before their surgery, patients were asked to complete a questionnaire that asked them to forecast their 2-week postoperative upper extremity disability and pain by completing a shortened DASH (QuickDASH) and pain VAS. At the 2-week postoperative visit, patients completed the QuickDASH and pain VAS to assess their realized disability and pain. They were not shown their forecasted scores. All patients were given a standardized preoperative education sheet as part of at their preoperative visit, which included information about the day of surgery, wound care, recommended over-the-counter medication as well as narcotic pain medication, and postoperative swelling. This sheet and the verbal counseling from the surgeon did not change during the course of the study
Exploratory variables collected included age, sex, socioeconomic elements (race-ethnicity, health insurance, years of education, work status, marital status, annual salary) (Table), QuickDASH, pain VAS, General Self Efficacy Scale [GSE-6], Pain Catastrophizing Scale [PCS-4], and Patient Health Questionnaire [PHQ-2]. Procedure type was also collected (Table ). Primary response variables were realized QuickDASH and VAS pain scores at 2 weeks postsurgery. Additionally, we assessed whether the forecasted disability score was within the minimal clinically important difference (MCID) of the realized disability score. The MCID depicts the smallest improvement in score to show a change that is clinically meaningful for a patient . We used a MCID value of 15.9 for the QuickDASH based on prior work . Written consent was obtained from all participants and completed questionnaires were electronically transcribed and complied using REDCap (Research Electronic Data Capture; Vanderbilt University, Nashville, TN, USA), a HIPAA-compliant, web-based application used to collect data for research purposes .
The QuickDASH, which is a validated shortened version of the DASH, was used to measure upper extremity-specific disability . It contains 11 items, which are answered on a 5-point Likert scale. The score is scaled to a value between 0 (no disability) and 100 (most severe disability). The instrument cannot be used if more than one question is missing. Mean preoperative QuickDASH score was 42.14 ± 28.30 and mean postoperative QuickDASH score was 44.85 ± 26.27.
The GSE-6 is a validated six-item test for the assessment of general self-efficacy . General self-efficacy is the belief that a person is able to control difficult environmental stresses by taking adaptive action. Participants are asked to rate the degree in which each item applies to them on a scale ranging from “not at all true” (1) to “exactly true” (4). Questions are given a value of 10 to 40, where a response of “not at all” corresponds to a value of 10, “hardly at all” a value of 20, etc. The final score is the average across all six questions. The total score ranges from 10 to 40 with a higher score indicating greater self-efficacy; patients had a mean GSE-6 score of 30.66 ± 7.1.
The PCS-4 is a validated tool that contains four items aimed at assessing the patients’ perception of their own pain . It measures the domains of rumination, magnification, and helplessness to determine the level of pain catastrophizing. Items are scored on a scale from 0 to 4. The final score is the sum with a low score of 0 points and a maximum possible score of 16 points, with a higher score indicating greater pain catastrophizing; patients had a mean PCS-4 score of 5.3 ± 3.8.
We performed an a priori sample size estimation based on a Pearson’s correlation. To detect a 0.30 correlation between forecasted disability and realized disability at an α of 0.05 and 80% power, a total sample size of 84 patients was calculated. Descriptive statistics were performed for each outcome of interest and shown as the frequency for categorical variables and mean (± SD) for continuous variables. Bivariate analysis was used to test the association of forecasted and realized disability and pain reporting Pearson’s correlation coefficients. We interpreted it using Evans’ classification: less than 0.20 is very weak, 0.20 to 0.39 is weak, 0.40 to 0.59 is moderate, 0.60 to 0.79 is strong and 0.80 or greater is a very strong correlation . Unpaired t-tests were performed to test the association of demographic variables (for example, men versus women) and the association of forecasted and realized disability and pain levels. One-way analysis of variance was used for variables with multiple groups (for example, annual salary, and ethnicity). All p values < 0.05 were considered statistically significant.
Forecasted postoperative disability was moderately correlated with realized postoperative disability (r = 0.59; p < 0.001), and forecasted pain was weakly correlated with realized pain (r = 0.28; p = 0.01) (Fig. ). Forecasted postoperative disability correlated with realized postoperative disability in patients who underwent surgery for trigger finger release (r = 0.87; p < 0.001) and fracture fixation (r = 0.49; p = 0.020), though not for carpal tunnel release (r = 0.29; p = 0.247) and mass excision (r = 0.22; p = 0.545). Forecasted postoperative pain was not correlated with realized pain in patients who underwent trigger finger release (r = 0.55; p = 0.080), fracture fixation (r = 0.40; p = 0.074), carpal tunnel release (r = 0.03; p = 0.257), and mass excision (r = 0.34; p = 0.337), although our study was not powered for this level of analysis. A total of 47% (n = 40) of patients were able to forecast their disability score within the MCID of their realized disability score (using an MCID of 15.9). Overall, 56% (n = 48) of all patients were within the MCID of their realized disability. The percent of patients within the MCID of their realized disability score in different procedures were: 64% of patients who underwent trigger finger release, 41% in fracture fixation, 56% in carpal tunnel release, and 90% in mass excision.
Older age was correlated with a greater difference between forecasted and realized disability (r = 0.27; p = 0.011). Older patients overestimated their disability score and had less disability postoperatively than what they forecasted (Table ). Greater catastrophic thinking, general self-efficacy, and depression did not correlate with a greater difference between forecasted and realized disability postoperatively. Increased symptoms of depression were correlated with a greater difference between forecasted and realized pain (r = -0.24; p = 0.028); this suggests that patients with symptoms of depression had greater realized pain postoperatively than what they forecasted preoperatively. Neither catastrophic thinking nor general self-efficacy was correlated with differences in forecasted and actual pain postoperatively. Increased symptoms of depression were correlated with greater realized disability postoperatively (r = 0.37; p < 0.001); and increased symptoms of depression (r = 0.42; p < 0.001) and catastrophic thinking (r = 0.31; p = 0.004) were correlated with greater realized pain postoperatively (Table ).
There is great interest in using PROMs such as disability and pain to assess quality in orthopaedic surgery, however, factors that can independently influence PROMs, like depression, continue to be identified. There is no single instrument that can assess all these factors to identify patients who may have a poor outcome after intervention. Prior work suggests a patient’s self-awareness of their own health can be predictive of their outcome [8, 18, 23]. Forecasting postoperative disability before orthopaedic surgery could be used to assess a patient’s expected disability and inform interventions, such as cognitive behavioral therapy, for those at risk for poor outcomes. We found that patients are able to moderately forecast their postoperative disability and weakly forecast their postoperative pain. Forecasting could be used during routine care to identify patients at risk of poor outcomes and inform preoperative discussions and behavioral interventions .
Our findings should be interpreted while considering the limitations of our study design. First, we studied a cohort of patients from a suburban academic hand surgery office of one surgeon. Although this study should be replicated in other populations, we are confident that the phenomenon of forecasting disability, that is, understanding one’s ability to cope with impairment, will be similar regardless of orthopaedic condition and geography. Since forecasting is a complex entity, factors beyond the psychosocial factors assessed in this study, such as the preoperative discussion, patient expectations, and prior experience with surgery could have affected a patient’s ability to forecast. For example, one study found that routine preoperative counseling may affect how patients experience pain and can decrease opioid consumption . To control for this potential confounder, we used a standardized protocol explaining surgery, the postoperative process, and pain medications, although this does not account for patient-specific discussions focused on postoperative function or other patient-level variables. The inclusion of only one surgeon also minimized variation in discussion that would have occurred with multiple surgeons. Although there was moderate correlation in forecasted disability scores (r = 0.59), there was only weak correlation in pain scores (r = 0.28). This suggests that forecasting disability could be informative as a preoperative tool, but not pain. Lastly, correlation does not imply causation. The information obtained from the forecasted disability score, regardless of why it is associated with the realized score, gives insight into the patient’s postoperative disability.
In our study, patients were able to forecast their postoperative disability and were also able to forecast their postoperative pain, although to a lesser degree than disability. This aligns with one previous study in orthopaedic surgery, which found a correlation between anticipated pain preoperatively and pain experienced postoperatively . These findings suggest that patients can use their own experiences and understanding of their ability to cope to predict their own postoperative outcomes. Our findings also strengthen the assertion that patients understand themselves well and thus may be able to accurately predict their own disability better than a risk-adjusted predictive model [19, 21, 38]. For example, almost half of our patients’ forecasted disability was within the MCID of their realized disability score. It is plausible that physicians could use preoperative forecasted PROMs measuring disability to risk-stratify patients who are forecasting greater disability than expected. For example, the state of Minnesota passed legislation that requires collection of PROMs before and after TKA and lumbar spine surgery . Similarly, Medicare includes functional assessment before and after total joint arthroplasty as a quality measure . As PROMs become increasingly integrated into routine care, forecasting disability could be a patient-level risk stratifying tool before orthopaedic intervention, although further analyses are needed .
Further, we found that psychosocial and demographic factors may influence a patients’ ability to forecast postoperative disability and pain. We found that symptoms of depression were associated with greater realized pain postoperatively and that increased symptoms of depression were correlated with a worse ability to forecast. Similarly, it has previously been shown that presurgery distress, depression, smoking, and less education are predictors of pain and postsurgical pain severity [4, 25, 29, 33, 34, 36]. Taken together, a more negative mindset may lead to greater pain after surgery. We also found that age, but no other demographic factors, were associated with a patient’s ability to forecast. This could be secondary to greater insight into one’s ability to cope with adversity based on previous experiences. Identifying modifiable risk factors associated with greater forecasted disability could potentially guide preoperative interventions for patients at risk for greater than expected disability. For example, a patient who forecasts substantial postoperative disability and is also a catastrophic thinker may benefit from preoperative cognitive behavioral therapy. In this way, forecasting may be a way to screen patients before surgery, and direct those at risk for a poor outcome into a care pathway that includes assessment of psychosocial factors that inform behavioral interventions. For instance, if payers begin requiring an improvement in PROM scores after surgery, forecasting allows providers to engage these patients in preoperative interventions, such as preoperative education, cognitive-behavioral therapy, or treatment modifications such as medication, diabetes control, better management of pain and depressive symptoms to improve patient outcomes [12, 30, 31, 39].
Before orthopaedic surgery, forecasted disability, and to a lesser degree forecasted pain, could be used as a screening tool that informs surgeons and health systems about patients at risk for poor patient-reported outcomes. This information could be used to direct at-risk patients toward a care pathway that potentially investigates and addresses underlying factors, such as depression, associated with forecasted disability that is greater than expected. Such an application would require condition-specific benchmarking of expected postoperative PROM scores, a process already underway . Further confirmatory studies are needed to assess the predictive ability of forecasting for other common orthopaedic conditions such as spine surgery or total joint arthroplasty. Exploration into condition-, patient-, and surgeon-specific variables that may influence the accuracy of forecasting that could be used to improve the predictive ability of those at risk for poor outcomes are also needed.