Using Unsupervised Machine Learning Methods to Cluster Comorbidities in a Population-Based Cohort of Patients With Rheumatoid Arthritis

OBJECTIVE

To identify clusters of comorbidities in patients with rheumatoid arthritis (RA) using 4 methods and to compare to patients without RA.

METHODS

In this retrospective, population-based study, residents of 8 Minnesota counties with prevalent RA as of January 1, 2015 were identified. Age-, sex-, and county-matched non-RA comparators were selected from the same underlying population. Diagnostic codes were retrieved for 5 years before January 1, 2015. Using 2 codes ≥30 days apart, 44 previously defined morbidities and 11 nonoverlapping chronic disease categories based on Clinical Classifications Software were defined. Unsupervised machine learning methods of interest included hierarchical clustering, factor analysis, K-means clustering, and network analysis.

RESULTS

Two groups of 1,643 patients with and without RA (72% female; mean age 63.1 years in both groups) were studied. Clustering of comorbidities revealed strong associations among mental/behavioral comorbidities and among cardiovascular risk factors and diseases. The clusters were associated with age and sex. Differences between the 4 clustering methods were driven by comorbidities that are rare and those that were weakly associated with other comorbidities. Common comorbidities tended to group together consistently across approaches. The instability of clusters when using different random seeds or bootstrap sampling impugns the usefulness and reliability of these methods. Clusters of common comorbidities between RA and non-RA cohorts were similar.

CONCLUSION

Despite the higher comorbidity burden in patients with RA compared to the general population, clustering comorbidities did not identify substantial differences in comorbidity patterns between the RA and non-RA cohorts. The instability of clustering methods suggests caution when interpreting clustering using 1 method.

Epidemiology and Trends in Survival of Systemic Sclerosis in Olmsted County (1980-2018): A Population-based Study

Background/Purpose

To update the epidemiology of systemic sclerosis (SSc) and evaluate the performance of the ACR/EULAR 2013 vs. 1980 ARA classification criteria in a U.S. population-based cohort.

Methods

An inception cohort of patients with incident SSc from January 1, 1980, through December 31, 2018, in Olmsted County, Minnesota, was identified based on comprehensive individual medical record review. Incidence and prevalence rates were age- and sex-adjusted to the 2010 US white population. Survival rates were compared with expected rates in the general population. Fulfillment of 1980 and 2013 classification criteria was ascertained.

Results

A total of 85 incident cases of SSc (91% female, mean age 55.4 ± 16 y) and 49 prevalent cases on Jan 1, 2015 were identified. The overall age- and sex-adjusted annual incidence was 25 (95% CI 20-31) per million population, with no change in incidence over time. The age- and sex-adjusted prevalence was 436 (95% CI: 313-558) per 1,000,000 population. 77 (91%) patients fulfilled the 2013 classification criteria, and 38 (45%) fulfilled the 1980 criteria. Mortality among patients with SSc was significantly higher in comparison to the general population, with a standardized mortality ratio of 2.48 (95% CI:1.76-3.39) and no evidence of improvement over time.

Conclusions

SSc developed in 25 persons/million/year with no change over the 39-year study period. The 2013 classification criteria perform significantly better than the 1980 criteria but failed to classify 9% of patients. SSc portends a 2.5-fold higher risk of mortality than the general population, with no evidence of improved survival over time.

Comprehensive assessment of multimorbidity burden in a population-based cohort of patients with rheumatoid arthritis

Objective

To comprehensively assess multimorbidity burden in patients with rheumatoid arthritis (RA) in order to unify the multimorbidity definition for RA research and clinical practice.

Methods

In this population-based study, residents of eight Minnesota counties with prevalent RA on 1 January 2015 were identified. Age, sex and county-matched non-RA comparators were selected from the same population. Diagnostic codes were retrieved for 5 years before 1 January 2015. Using two codes ≥30 days apart, 44 previously defined morbidities and 78 non-overlapping chronic disease categories based on Clinical Classification Software were defined. Prevalence of each morbidity in the RA versus non-RA cohorts was compared using false discovery rate to adjust for multiple comparisons. Morbidities more common in RA than non-RA and those with prevalence ≥5% were retained.

Results

1643 patients with RA and 1643 non-RA subjects (72% women; mean age 63.1 years) were studied. Using the 44 morbidities, multimorbidity (defined as 2+ morbidities) was present in 1411 (86%) of RA and 1164 (71%) of non-RA subjects (p<0.001) with 5+ morbidities present in 907 (55%) of RA and 619 (38%) of non-RA (p<0.001). Patients with RA had significantly higher prevalence of 24 of the 44 morbidities compared with non-RA, especially interstitial lung disease, fibromyalgia, osteoarthritis and osteoporosis. Among the additional 78 categories, 7 were significantly higher in RA than non-RA, including organic sleep disorders, vitamin D deficiency and foot ulcers.

Conclusion

Patients with RA have a higher prevalence of multimorbidity compared with non-RA subjects. These results confirm the list of 44 morbidities and add several other morbidities of interest in RA.

Outpatient healthcare utilization among incident cases of systemic sclerosis: results from a population-based US cohort (1988-2016)

OBJECTIVE

Systemic sclerosis (SSc) is a complex, heterogeneous connective tissue disease with multiorgan dysfunction. This study aimed to compare healthcare utilization among incident cases of SSc versus age- and gender-matched comparators.

METHOD

A population-based cohort of physician-diagnosed patients with SSc in Olmsted County, MN, USA, from 1 January 1988 to 31 December 2016 was assembled. A 2:1 cohort of age- and gender-matched non-SSc subjects was randomly selected for comparison. Patients were followed until death, migration from Olmsted County, or 31 December 2017. Outpatient utilization data were obtained beginning 12 months before the SSc incidence/index date and compared using negative binomial and multinomial models. Services were summarized as visit-days to avoid overestimation of services provided.

RESULTS

The study included 69 incident SSc cases and 138 non-SSc comparators (mean ± sd age 57 ± 16 years at diagnosis/index, 90% female). Patients with SSc had higher utilization of outpatient physician, laboratory, and combined radiology visit-days annually for the year before and for each of the first 5 years after diagnosis than comparators. Among patients with SSc, healthcare utilization was highest during the year of SSc diagnosis. Rate ratios comparing utilization in patients with and without SSc ranged from 1.8 to 3.0 for all comparisons.

CONCLUSION

Higher utilization of outpatient physician, laboratory, and radiology visit-days was observed among patients with SSc compared to non-SSc subjects throughout 5 years of disease duration, indicating high and continued care needs in this patient population. The highest utilization of services among SSc patients occurred during the year of SSc diagnosis.

Hospitalization Rates Are Highest in the First 5 Years of Systemic Sclerosis: Results From a Population-based Cohort (1980-2016)

OBJECTIVE

Few studies have estimated the healthcare resource usage of patients with systemic sclerosis (SSc). The purpose of this study was to compare hospitalization among incident cases of SSc vs age- and sex-matched comparators.

METHODS

A retrospective, population-based cohort of patients with SSc in Olmsted County, Minnesota, from January 1, 1980, to December 31, 2016, was assembled. A 2:1 cohort of age- and sex-matched patients without SSc from the same population was randomly selected for comparison. All hospitalizations in the geographic area from January 1, 1987, to September 30, 2018, were obtained. Rates of hospitalization, lengths of stay, and readmissions were compared between groups.

RESULTS

There were 76 incident SSc cases and 155 non-SSc comparators (mean age 56 ± 16 yrs at diagnosis/index, 91% female) included. Rates of hospitalization among cases and comparators were 31.9 and 17.9 per 100 person-years, respectively (rate ratio [RR] 1.78, 95% CI 1.52-2.08). Hospitalization rates were higher in patients with SSc than comparators during the first 5 years after SSc diagnosis (RR 2.16, 95% CI 1.70-2.74). This difference decreased over time and was no longer significant at ≥ 15 years after SSc incidence/index. Lengths of stay (median [IQR] 4 [2-6] vs 3 [2-6], P = 0.52) and readmission rates (25% vs 23%, P = 0.51) were similar between groups.

CONCLUSION

Patients with SSc were hospitalized more frequently than comparators, indicating high inpatient care needs in this population. Hospitalization rates were highest during the first 5 years following SSc diagnosis.

Evidence of Diagnostic and Treatment Delay in Seronegative Rheumatoid Arthritis: Missing the Window of Opportunity

OBJECTIVES

To compare the time from first joint swelling to fulfillment of the American College of Rheumatology/European League Against Rheumatism classification criteria between patients with seropositive and seronegative rheumatoid arthritis (RA) and to assess the impact of seronegative status on the time from first joint swelling to initiation of disease-modifying antirheumatic drug (DMARD) therapy and achievement of remission.

PATIENTS AND METHODS

Times from first provider-documented joint swelling to fulfillment of the 1987 and 2010 American College of Rheumatology/European League Against Rheumatism criteria and to the clinical diagnosis of RA were measured in a population-based cohort of adults with incident RA between January 1, 2009, and December 31, 2014. Disease characteristics and achievement of remission were compared between seropositive (rheumatoid factor positive and/or anti-citrullinated peptide antibody positive) and seronegative (rheumatoid factor negative/anti-citrullinated peptide antibody negative) patients.

RESULTS

The median time from first joint swelling to fulfillment of the 1987 (48 [interquartile range (IQR), 0-300] days vs 2 [IQR, 0-45] days; P=.001) and 2010 (14 [IQR, 0-196] days vs 0 [IQR, 0-29] days; P=.004) classification criteria and the median time from first joint swelling to the clinical diagnosis of RA (187 [IQR, 13-503] days vs 11 [IQR, 0-76] days; P<.001) were significantly longer in seronegative patients than in seropositive patients. The median time from first joint swelling to first prescribed DMARD therapy was significantly longer in seronegative patients (40 [IQR, 5-199] days vs 14 [IQR, 0-73] days; P=.01). Patients with seronegative RA were less likely to achieve remission (28% vs 50% at 5 years after fulfillment of the 2010 criteria; P=.007), but there was no difference when the patient global score was removed from the remission definition.

CONCLUSION

Patients with seronegative RA experienced a delay in diagnosis, according to both the 1987 and 2010 classification criteria, as well as a delay in the initiation of DMARD therapy. Patients with seronegative RA were also less likely to attain remission, suggesting that the window of opportunity for intervention may be more frequently missed in this group.

Rituximab Therapy for Systemic Rheumatoid Vasculitis: Indications, Outcomes, and Adverse Events

Objective.

To characterize the indication, outcomes, and adverse effects of rituximab (RTX) treatment in a large single-center cohort of patients with systemic rheumatoid vasculitis (RV).

Methods.

We retrospectively reviewed the medical charts of 17 patients treated with RTX for systemic RV from 2000 to 2017. Clinical characteristics, outcomes, and adverse effects were analyzed.

Results.

At RV diagnosis, mean age was 59 years, 59% were female, 94% were white, and 82% had positive rheumatoid factor. At the time of initiating RTX, median Birmingham Vasculitis Activity Score for rheumatoid arthritis was 4.0 (interquartile range 2.0–7.5). RV presented in the skin in 8 patients (47%), as mononeuritis multiplex in 2 (12%), inflammatory ocular disease in 2 (12%), and affected multiple organ systems in 5 (29%). RTX was used for induction therapy in 8 patients (47%), relapsing RV in 4 (24%), second-line therapy in 2 (12%), and salvage therapy or in combination with another agent in 3 (18%). At 3 months, 2 (13%) of 15 patients with available followup information achieved complete remission (CR), and 10 (67%) achieved partial response (PR). At 6 months, 6 patients (40%) achieved CR, 8 (53%) achieved PR, and one had no response. At 12 months, 8 of 13 patients with available records (62%) had CR and 5 patients (38%) had PR.

Conclusion.

Systemic RV is difficult to treat effectively. CR of RV was achieved in 62% and PR in 38% of patients within 12 months of RTX use. Further evidence is needed to inform treatment for patients with RV.

Breast-specific gamma imaging as an adjunct modality for the diagnosis of invasive breast cancer with correlation to tumour size and grade

OBJECTIVES

The purpose of this study was to determine the sensitivity of breast-specific gamma imaging (BSGI) in the detection of invasive breast cancers and to characterise the sensitivity of BSGI based on tumour size and pathological grade.

METHODS

139 females with invasive carcinoma who underwent BSGI were retrospectively reviewed. Patients were injected in the antecubital vein with 20-30 mCi (925-1110 MBq) of (99m)Tc-sestamibi. Images were obtained with a high-resolution, breast-specific gamma camera (Dilon 6800; Dilon Technologies, Newport News, VA) and were categorised based on radiotracer uptake as normal, normal with heterogeneous uptake, probably abnormal and abnormal. For a positive examination, the region of the area of increased uptake had to correlate with the laterality and location of the biopsy-proven cancer.

RESULTS

149 invasive cancers in 139 patients with a mean size of 1.8 cm (0.2-8.5 cm) were included. 146 were identified with BSGI (98.0%). All cancers which measured ≥ 0.7 cm (n = 123) as well as all cancers grade 2 or higher (n = 102), regardless of tumour size, were identified with BSGI (100%). There were 6 cancers that were pathological grade 1 and measured <7 mm, of which 50% (3/6) were identified with BSGI. The overall sensitivity of BSGI for the detection of invasive breast cancer is 98.0%. The sensitivity for subcentimetre cancers is 88.5% (23/26).

CONCLUSION

BSGI has a high sensitivity for the detection of invasive breast cancer. Our results demonstrate that BSGI detected all invasive breast cancers pathological grade 2 and higher regardless of size and all cancers which measured ≥ 7 mm regardless of grade. BSGI can reliably detect invasive breast cancers and is a useful adjunct imaging modality for the diagnosis of breast cancer.

Effects of an educational intervention for general practitioners in adolescent health care principles: a randomized controlled study

Objective To evaluate the effectiveness of an educational intervention in adolescent health designed for general practitioners, in accordance with evidence-based practice in continuing medical education. Design Randomized, controlled trial with baseline testing and 7- and 13-month follow-ups. Setting The intervention was delivered in local community settings to general practitioners in metropolitan Melbourne, Australia. Participants A total of 108 self-selected general practitioners. Intervention A multifaceted educational program (2.5 hours per week for 6 weeks) in the principles of adolescent health care, followed 6 weeks later by a 2-hour session of case discussion and debriefing. Outcome measures Objective ratings of videotaped consultations with standardized adolescent patients and self-completion questionnaires were used to measure general practitioners' knowledge, skill, and self-perceived competency; satisfaction with the program; and self-reported change in practice. Results 103 of 108 physicians (95%) completed all phases of the intervention and evaluation protocol. The intervention group showed significantly greater improvements than the control group in all outcomes at the 7-month follow-up (all subjects P<0.03), except for the standardized patients' rating of rapport and satisfaction (P=0.12). 104 participants (96%) found the program appropriate and relevant. At the 13-month follow-up, most improvements were sustained, the standardized patients' rating of confidentiality fell slightly, and the objective assessment of competence further improved. 106 physicians (98%) reported a change in practice attributable to the intervention. Conclusions General practitioners were willing to complete continuing medical education in adolescent health and its evaluation. The design of the intervention, using evidence-based educational strategies, proved effective and expeditious in achieving sustainable and large improvements in knowledge, skill, and self-perceived competency.

Evaluation of the effectiveness of an educational intervention for general practitioners in adolescent health care: randomised controlled trial

OBJECTIVE

To evaluate the effectiveness of an educational intervention in adolescent health designed for general practitioners in accordance with evidence based practice in continuing medical education.

DESIGN

Randomised controlled trial with baseline testing and follow up at seven and 13 months.

SETTING

Local communities in metropolitan Melbourne, Australia.

PARTICIPANTS

108 self selected general practitioners.

INTERVENTION

A multifaceted educational programme for 2.5 hours a week over six weeks on the principles of adolescent health care followed six weeks later by a two hour session of case discussion and debriefing.

OUTCOME MEASURES

Objective ratings of consultations with standardised adolescent patients recorded on videotape. Questionnaires completed by the general practitioners were used to measure their knowledge, skill, and self perceived competency, satisfaction with the programme, and self reported change in practice.

RESULTS

103 of 108 (95%) doctors completed all phases of the intervention and evaluation protocol. The intervention group showed significantly greater improvements in all outcomes than the control group at the seven month follow up except for the rapport and satisfaction rating by the standardised patients. 104 (96%) participants found the programme appropriate and relevant. At the 13 month follow up most improvements were sustained, the confidentiality rating by the standardised patients decreased slightly, and the objective assessment of competence further improved. 106 (98%) participants reported a change in practice attributable to the intervention.

CONCLUSIONS

General practitioners were willing to complete continuing medical education in adolescent health care and its evaluation. The design of the intervention using evidence based educational strategies proved an effective and quick way to achieve sustainable and large improvements in knowledge, skill, and self perceived competency.

Clinical significance of anismus in encopresis

Treatments designed to relieve paradoxical contraction of the anal sphincters during defecation (anismus) have had limited success in children with encopresis. This has raised doubts as to the clinical relevance of this diagnosis in childhood as anorectal dysfunction. Our aim was to determine whether, in patients who had treatment-resistant encopresis, the presence of electromyographic anismus was associated with increased faecal retention. Sixty-eight children with soiling (mean age 8.7+/-2.06 years) were assessed by clinical examination, abdominal radiography and then with anorectal manometry. Patients with electromyographic anismus (n=32; 47%) had significantly increased radiographic rectal faecal retention and were significantly less likely to be able to defecate water-filled balloons. There were no significant differences in response to prior therapy, history of primary encopresis, behavioural adjustment or in sociodemographic data. Our results suggest that electromyographic anismus is associated with obstructed defecation and faecal retention.

Barriers to effective primary health care for adolescents

OBJECTIVES

To assess general practitioners' (GPs') perceptions of barriers in the health care system that hinder provision of effective adolescent health care; and to assess the training needs of GPs for a future education program in adolescent health issues.

DESIGN

Retrospective questionnaire survey.

METHODS

A 66-item questionnaire was mailed during May and June 1995 to a random stratified sample of 997 rural and urban Victorian GPs. A Practice Assessment Task enabling respondents to fulfil Royal Australian College of General Practitioners' quality assurance requirements was mailed on return of completed questionnaires if requested.

RESULTS

The response rate was 72.4%. After adjusting for stratified sampling, 77% (95% confidence interval [CI], 73%-81%) of respondents favoured individual Medicare cards for adolescents from the age of 16. Thirty-eight per cent (95% CI, 34%-43%) were less willing to charge for longer consultations because they feared Health Insurance Commission (HIC) investigation. Most respondents indicated that their undergraduate training in adolescent mental health issues was inadequate and 64% (95% CI, 61%-68%) found it difficult to obtain advice about complex mental health problems. An interest in continuing medical education in adolescent health issues was expressed by 82% (95% CI, 79%-86%) of respondents.

CONCLUSIONS

Individual Medicare cards should be automatically issued to adolescents from the age of 16 to improve their ability to access health care. Revision of the Medicare rebate system and clarification of HIC's investigative functions may improve GP's effectiveness in adolescent consultations. Undergraduate medical and GP training should include curriculum relevant to adolescent health care.

Fecal incontinence after the surgical treatment of Hirschsprung disease

We examined 60 children 8.9 years (+/- 2.6 years) after surgical treatment of Hirschsprung disease to determine the extent of fecal incontinence. Thirty-two children (53%) had significant fecal soiling and 16 (27%) less severe soiling. The prevalence of incontinence did not diminish with increasing age.