Development of prediction models for clinically meaningful improvement in PROMIS scores after lumbar decompression

BACKGROUND

The ability to preoperatively predict which patients will achieve a minimal clinically important difference (MCID) after lumbar spine decompression surgery can help determine the appropriateness and timing of surgery. Patient-Reported Outcome Measurement Information System (PROMIS) scores are an increasingly popular outcome instrument.

PURPOSE

The purpose of this study was to develop algorithms predictive of achieving MCID after primary lumbar decompression surgery.

PATIENT SAMPLE

This was a retrospective study at two academic medical centers and three community medical centers including adult patients 18 years or older undergoing one or two level posterior decompression for lumbar disc herniation or lumbar spinal stenosis between January 1, 2016 and April 1, 2019.

OUTCOME MEASURES

The primary outcome, MCID, was defined using distribution-based methods as one half the standard deviation of postoperative patient-reported outcomes (PROMIS physical function, pain interference, pain intensity).

METHODS

Five machine learning algorithms were developed to predict MCID on these surveys and assessed by discrimination, calibration, Brier score, and decision curve analysis. The final model was incorporated into an open access digital application.

RESULTS

Overall, 906 patients completed at least one PROMs survey in the 90 days before surgery and at least one PROMs survey in the year after surgery. Attainment of MCID during the study period by PROMIS instrument was 74.3% for physical function, 75.8% for pain interference, and 79.2% for pain intensity. Factors identified for preoperative prediction of MCID attainment on these outcomes included preoperative PROs, percent unemployment in neighborhood of residence, comorbidities, body mass index, private insurance, preoperative opioid use, surgery for disc herniation, and federal poverty level in neighborhood of residence. The discrimination (c-statistic) of the final algorithms for these outcomes was 0.79 for physical function, 0.74 for pain interference, and 0.69 for pain intensity with good calibration. The open access digital application for these algorithms can be found here: https://sorg-apps.shinyapps.io/promis_pld_mcid/ CONCLUSION: Lower preoperative PROMIS scores, fewer comorbidities, and certain sociodemographic factors increase the likelihood of achieving MCID for PROMIS after lumbar spine decompression.

Surgeon-level variance in achieving clinical improvement after lumbar decompression: the importance of adequate risk adjustment

BACKGROUND CONTEXT

Patient-Reported Outcome Measurement Information System (PROMIS) scores are increasingly utilized in clinical care. However, it is unclear if PROMIS can discriminate surgeon performance on an individual level.

PURPOSE

The purpose of this study was to examine surgeon-level variance in rates of achieving minimal clinically important difference (MCID) after lumbar decompression.

PATIENT SAMPLE

This is a prospective, observational cohort study performed across a healthcare enterprise (two academic medical centers and three community centers). Patients 18 years or older undergoing one- to two-level primary decompression for lumbar disc herniation (LDH) or lumbar spinal stenosis (LSS) were included.

OUTCOME MEASURES

The primary outcome was achievement of MCID, using a distribution-based method, on paired PROMIS physical function scores.

METHODS

Descriptive statistics were generated to examine the baseline characteristics of the study cohort. Bivariate analyses were used to examine the impact of surgeon-level variance on rates of MCID. Multivariable analyses were used to examine the risk-adjusted impact of surgeon-level variance on rates of MCID.

RESULTS

Overall, 636 patients treated by nine surgeons were included. The median patient age was 58 [interquartile range (IQR): 46-70] and 62.3% (n=396) were female. Among all patients, 56.9% (n=362) underwent surgery for LDH. The overall rate of achieving MCID was 75.8% (n=482). Of the surgeons, the median years in practice were 12 (range 4-31) and 55.6% (n=5) were in academic practice settings. On bivariate analysis, patients treated by one of the surgeons had lower rates of achieving MICD (odds ratio=0.37, 95% confidence interval: 0.15-0.91, p=.03). However, on multivariable analysis adjusting for operative indication (LDH vs. LSS), body mass index, number of comorbidities, percent unemployment in patient zip code, and preoperative PROMIS physical function scores, all surgeons were equally likely to obtain MCID.

CONCLUSIONS

In this cohort, variance in PROMIS scores after primary lumbar decompression is influenced by patient-related factors and not by individual surgeon. Adequate risk adjustment is needed if ascertaining clinical improvement on an individual surgeon basis.

LEVEL OF EVIDENCE

2.

Can natural language processing provide accurate, automated reporting of wound infection requiring reoperation after lumbar discectomy?

BACKGROUND

Surgical site infections are a major driver of morbidity and increased costs in the postoperative period after spine surgery. Current tools for surveillance of these adverse events rely on prospective clinical tracking, manual retrospective chart review, or administrative procedural and diagnosis codes.

PURPOSE

The purpose of this study was to develop natural language processing (NLP) algorithms for automated reporting of postoperative wound infection requiring reoperation after lumbar discectomy.

PATIENT SAMPLE

Adult patients undergoing discectomy at two academic and three community medical centers between January 1, 2000 and July 31, 2019 for lumbar disc herniation.

OUTCOME MEASURES

Reoperation for wound infection within 90 days after surgery METHODS: Free-text notes of patients who underwent surgery from January 1, 2000 to December 31, 2015 were used for algorithm training. Free-text notes of patients who underwent surgery after January 1, 2016 were used for algorithm testing. Manual chart review was used to label which patients had reoperation for wound infection. An extreme gradient-boosting NLP algorithm was developed to detect reoperation for postoperative wound infection.

RESULTS

Overall, 5,860 patients were included in this study and 62 (1.1%) had a reoperation for wound infection. In patients who underwent surgery after January 1, 2016 (n=1,377), the NLP algorithm detected 15 of the 16 patients (sensitivity=0.94) who had reoperation for infection. In comparison, current procedural terminology and international classification of disease codes detected 12 of these 16 patients (sensitivity=0.75). At a threshold of 0.05, the NLP algorithm had positive predictive value of 0.83 and F1-score of 0.88.

CONCLUSION

Temporal validation of the algorithm developed in this study demonstrates a proof-of-concept application of NLP for automated reporting of adverse events after spine surgery. Adapting this methodology for other procedures and outcomes in spine and orthopedics has the potential to dramatically improve and automatize quality and safety reporting.

Algorithmic approach to the management of the patient with lumbar spinal stenosis

A 71-year-old generally healthy woman presents for her first visit in 3 years. She ambulates slowly from the waiting room, with a more stooped posture than previously. She reports a 2-year history of slowly worsening buttock and leg pain when she walks any distance. She has noticed that her symptoms are much less when she leans on a shopping cart in the grocery store. Her buttock/leg pain resolves within a few minutes when she sits down. The patient exhibits signs and symptoms suggestive of lumbar spinal stenosis.

Primary intracerebral Hodgkin's lymphoma

The case is reported of a 51 year old man with primary intracerebral Hodgkin's lymphoma treated by surgical excision, intrathecal chemotherapy and whole-brain irradiation. One year later the patient had no evidence of Hodgkin's lymphoma intracranially or elsewhere. The possible histogenesis of this rare condition is discussed and a brief review of the literature is presented.

Carotid ultrasonic arteriography combined with real time spectral analysis: a comparison with angiography

In a one year period, 400 patients were examined by carotid ultrasonic arteriography combined with real time sound spectral analysis. Contrast arteriography was performed on 112 patients. The noninvasive examinations and contrast arteriograms of these patients were reviewed by two independent observers, who were unaware of the findings of the other diagnostic modality. There were 190 vessels available for comparison. The internal carotid artery was classified as normal, less than 50% stenosis, greater than or equal to 50% stenosis, or occluded. The specificity of these non-invasive tests was 79%. The sensitivity in correctly identifying stenoses as less than 50% or greater than or equal to 50 was 85% and 91% respectively. The sensitivity in identifying occluded vessels was 90%. The positive and negative predictive values were 87% and 92% respectively. Noninvasive evaluation correctly identified six lesions misdiagnosed by arteriography. We found pulsed Doppler arteriography combined with real time spectral analysis to be an accurate, direct, noninvasive method of assessing the extracranial carotid arteries.