Impact of a Machine Learning-Based Decision Support System for Urinary Tract Infections: Prospective Observational Study in 36 Primary Care Practices.

BACKGROUND: There is increasing attention on machine learning (ML)-based clinical decision support systems (CDSS), but their added value and pitfalls are very rarely evaluated in clinical practice. We implemented a CDSS to aid general practitioners (GPs) in treating patients with urinary tract infections (UTIs), which are a significant health burden worldwide. OBJECTIVE: This study aims to prospectively assess the impact of this CDSS on treatment success and change in antibiotic prescription behavior of the physician. In doing so, we hope to identify drivers and obstacles that positively impact the quality of health care practice with ML. METHODS: The CDSS was developed by Pacmed, Nivel, and Leiden University Medical Center (LUMC). The CDSS presents the expected outcomes of treatments, using interpretable decision trees as ML classifiers. Treatment success was defined as a subsequent period of 28 days during which no new antibiotic treatment for UTI was needed. In this prospective observational study, 36 primary care practices used the software for 4 months. Furthermore, 29 control practices were identified using propensity score-matching. All analyses were performed using electronic health records from the Nivel Primary Care Database. Patients for whom the software was used were identified in the Nivel database by sequential matching using CDSS use data. We compared the proportion of successful treatments before and during the study within the treatment arm. The same analysis was performed for the control practices and the patient subgroup the software was definitely used for. All analyses, including that of physicians' prescription behavior, were statistically tested using 2-sided z tests with an α level of .05. RESULTS: In the treatment practices, 4998 observations were included before and 3422 observations (of 2423 unique patients) were included during the implementation period. In the control practices, 5044 observations were included before and 3360 observations were included during the implementation period. The proportion of successful treatments increased significantly from 75% to 80% in treatment practices (z=5.47, P<.001). No significant difference was detected in control practices (76% before and 76% during the pilot, z=0.02; P=.98). Of the 2423 patients, we identified 734 (30.29%) in the CDSS use database in the Nivel database. For these patients, the proportion of successful treatments during the study was 83%-a statistically significant difference, with 75% of successful treatments before the study in the treatment practices (z=4.95; P<.001). CONCLUSIONS: The introduction of the CDSS as an intervention in the 36 treatment practices was associated with a statistically significant improvement in treatment success. We excluded temporal effects and validated the results with the subgroup analysis in patients for whom we were certain that the software was used. This study shows important strengths and points of attention for the development and implementation of an ML-based CDSS in clinical practice. TRIAL REGISTRATION: ClinicalTrials.gov NCT04408976; https://clinicaltrials.gov/ct2/show/NCT04408976.

Relationship between trust and patient involvement in medical decision-making: A cross-sectional study.

INTRODUCTION: Patients vary in their preferences regarding involvement in medical decision-making. Current research does not provide complete explanation for this observed variation. Patient involvement in medical decision-making has been found to be influenced by various mechanisms, one of which could be patients' trust in physicians. The aim of this study was to examine whether trust in physicians fosters or impairs patient involvement in medical decision-making. This study also aimed to determine to what extent the relationship between trust and preferences regarding decision-making roles was influenced by the sociodemographic characteristics of the patients. We hypothesised that trust can both foster and impair patient involvement in medical decision-making. MATERIALS AND METHODS: A survey was sent out to members of the Nivel Dutch Health Care Consumer Panel in February 2016 (response rate = 47%, N = 703). The Wake Forest Physician Trust Scale was used to measure trust. Patient involvement was measured using two items based on the study published by Flynn and colleagues in 2006. Multiple regression analysis was used to analyse the relationship between trust and patient involvement. RESULTS: We found a negative relationship between trust and patient involvement in medical decision-making in men. Women with high trust reported to be more involved in medical decision-making compared to men with high trust. CONCLUSION: The results suggest that trust impairs involvement in medical decision-making for men but not for women. Further research could provide a more comprehensive explanation of the variation in patient preferences regarding involvement in medical decision-making to further elucidate which underlying mechanisms could enhance patient participation.

Guideline Adherence in Antibiotic Prescribing to Patients with Respiratory Diseases in Primary Care: Prevalence and Practice Variation.

Respiratory tract infections (RTIs) account for a large part of antibiotic prescriptions in primary care. However, guidelines advise restrictive antibiotic prescribing for RTIs. Only in certain circumstances, depending on, e.g., comorbidity, are antibiotics indicated. Most studies on guideline adherence do not account for this. We aimed to assess guideline adherence for antibiotic prescribing for RTIs as well as its variation between general practices (GPs), accounting for patient characteristics. We used data from electronic health records of GPs in the Netherlands. We selected patients who consulted their GP for acute cough, rhinitis, rhinosinusitis or sore throat in 2014. For each disease episode we assessed whether, according to the GP guideline, there was an indication for antibiotics, using the patient's sociodemographic characteristics, comorbidity and co-medication. We assessed antibiotic prescribing for episodes with no or an unsure indication according to the guidelines. We analysed 248,896 episodes. Diagnoses with high rates of antibiotic prescribing when there was no indication include acute tonsillitis (57%), strep throat (56%), acute bronchitis (51%) and acute sinusitis (48%). Prescribing rates vary greatly between diagnoses and practices. Reduction of inappropriate antibiotic prescribing remains a key target to tackle antimicrobial resistance. Insight into reasons for guideline non-adherence may guide successful implementation of the variety of interventions already available for GPs and patients.

Does shared decision-making reduce antibiotic prescribing in primary care?

Background: Increasing antibiotic resistance is recognized as a major threat to global health and is related to antibiotic prescription rates in primary care. Shared decision-making (SDM), the process in which patients and doctors participate together in making decisions, is argued to possibly promote more appropriate use of antibiotics and reduce prescribing. However, it is unknown whether in practice fewer antibiotics are prescribed where more SDM takes place. Objectives: To investigate whether more SDM is related to less antibiotic prescribing and whether this relationship differs between subgroups of patients (male/female and age groups). Patients and methods: A questionnaire survey was conducted among 2670 members of the Dutch Health Care Consumer Panel to measure SDM (response rate 45%). Average practice-level SDM scores were calculated for 15 general practices. Data from routine electronic health records of 8192 adult patients of these general practices participating in the Nivel Primary Care Database were used to assess relevant illness episodes (acute cough, acute rhinosinusitis and urinary tract infection), the indication for antibiotics and antibiotic prescriptions. Logistic multilevel regression analyses were performed to investigate the relationship between practice-level SDM and patient-level antibiotic prescriptions. Results: In practices where more SDM takes place, general practitioners prescribed fewer antibiotics for adult patients under the age of 40 years in preference-sensitive situations (i.e. situations in which antibiotics could be considered according to clinical guidelines). Conclusions: SDM can be a framework to reduce the prescribing of antibiotics and thus to control antibiotic resistance.

Is there a conflict between general practitioners applying guidelines for antibiotic prescribing and including their patients' preferences?

OBJECTIVES: One perceived barrier to guideline adherence is the existence of conflicting patient preferences. We examined whether patient preferences influence the prescription of antibiotics in general practice, and how this affects guideline adherence. We hypothesized that preferences play a larger role in prescribing antibiotics if the guideline allows for preferences to be taken into account, ie, if prescribing antibiotics is an option which can be considered rather than a clear recommendation to prescribe or not. We included three guidelines: acute cough, acute rhinosinusitis, and urinary tract infections. METHODS: Data from NIVEL (the Netherlands Institute for Health Services Research) Primary Care Database (NIVEL-PCD) were used to assess antibiotic indications and prescriptions. These data were combined with a questionnaire among members of NIVEL's Dutch Health Care Consumer Panel to examine patient preferences. According to NIVEL-PCD, 286 of these members contacted their general practitioner (GP) in 2015 for acute cough, acute rhinosinusitis or urinary tract infections. A logistic multilevel regression analysis was performed to test our hypothesis. RESULTS: Patient preferences do play a role in GPs' prescribing of antibiotics only in situations where, in accordance with the guideline, their use is an option which could be considered (interaction between indication and preference: p=0.049). If patients ask for antibiotics themselves in such situations, then GPs prescribe antibiotics more often. CONCLUSION: Patient preferences only play a role if the guideline provides room to take preferences into account. Therefore, our results do not suggest a conflict between applying guidelines and including patient preferences. Further research is recommended to examine this possible conflict in other situations.

Reconstructing spectral cues for sound localization from responses to rippled noise stimuli.

Human sound localization in the mid-saggital plane (elevation) relies on an analysis of the idiosyncratic spectral shape cues provided by the head and pinnae. However, because the actual free-field stimulus spectrum is a-priori unknown to the auditory system, the problem of extracting the elevation angle from the sensory spectrum is ill-posed. Here we test different spectral localization models by eliciting head movements toward broad-band noise stimuli with randomly shaped, rippled amplitude spectra emanating from a speaker at a fixed location, while varying the ripple bandwidth between 1.5 and 5.0 cycles/octave. Six listeners participated in the experiments. From the distributions of localization responses toward the individual stimuli, we estimated the listeners' spectral-shape cues underlying their elevation percepts, by applying maximum-likelihood estimation. The reconstructed spectral cues resulted to be invariant to the considerable variation in ripple bandwidth, and for each listener they had a remarkable resemblance to the idiosyncratic head-related transfer functions (HRTFs). These results are not in line with models that rely on the detection of a single peak or notch in the amplitude spectrum, nor with a local analysis of first- and second-order spectral derivatives. Instead, our data support a model in which the auditory system performs a cross-correlation between the sensory input at the eardrum-auditory nerve, and stored representations of HRTF spectral shapes, to extract the perceived elevation angle.

Increased cost sharing and changes in noncompliance with specialty referrals in The Netherlands.

INTRODUCTION: The compulsory deductible, a form of patient cost-sharing in the Netherlands, has more than doubled during the past years. There are indications that as a result, refraining from medical care has increased. We studied the relation between patient cost-sharing and refraining from medical care by evaluating noncompliance with referrals to medical specialists over several years. METHODS: Noncompliance with specialty referrals was assessed in the Netherlands from 2008 until 2013, using routinely recorded referrals from general practitioners to medical specialists and claims from medical specialists to health insurers. Associations with patient characteristics were estimated using multilevel logistic regression analyses. RESULTS: Noncompliance rates were approximately stable from 2008 to 2010 and increased from 18% in 2010 to 27% in 2013. Noncompliance was highest in adults aged 25-39 years. The increase was highest in children and patients with chronic diseases. No significantly higher increase among patients from urban deprived areas was found. DISCUSSION/CONCLUSION: Noncompliance increased during the rise of the compulsory deductible. Our results do not suggest a one-to-one relationship between increased patient cost-sharing and noncompliance with specialty referrals. In order to develop effective policy for reducing noncompliance, it is advisable to focus on the mechanisms for noncompliance in the groups with the highest noncompliance rates (young adults) and with the highest increase in noncompliance (children and patients with chronic diseases).

Measuring Temporal Resolution (Release of Masking) with a Hughson-Westlake Up-Down Instead of a Békèsy-Tracking Procedure.

BACKGROUND: A temporal resolution test in addition to the pure-tone audiogram may be of great clinical interest because of its relevance in speech perception and expected relevance in hearing aid fitting. Larsby and Arlinger developed an appropriate clinical test, but this test uses a Békèsy-tracking procedure for estimating masked thresholds in stationary and interrupted noise to assess release of masking (RoM) for temporal resolution. Generally the Hughson-Westlake up-down procedure is used in the clinic to measure the pure-tone thresholds in quiet. A uniform approach will facilitate clinical application and might be appropriate for RoM measurements as well. Because there is no golden standard for measuring the RoM in the clinic, we examine in the present study the Hughson-Westlake up-down procedure to measure the RoM and compare the results with the Békèsy-tracking procedure. PURPOSE: The purpose of the current study was to examine the differences between a Békèsy-tracking procedure and the Hughson-Westlake up-down procedure for estimating masked thresholds in stationary and interrupted noise to assess RoM. RESEARCH DESIGN: RoM is assessed in eight normal-hearing (NH) and ten hearing-impaired (HI) listeners through both methods. Results from both methods are compared with each other and with predicted thresholds from a model. DATA ANALYSIS: Wilcoxon signed-rank tests, paired t tests. RESULTS: Some differences between the two methods were found. We used a model to quantify the results of the two measurement procedures. The results of the Hughson-Westlake procedure were clearly better in agreement with the model than the results of the Békèsy-tracking procedure. Furthermore, the Békèsy-tracking procedure showed more spread in the results of the NH listeners than the Hughson-Westlake procedure. CONCLUSIONS: The Hughson-Westlake procedure seems to be an applicable alternative for measuring RoM for temporal resolution in the clinical audiological practice.

Evaluation of the preliminary auditory profile test battery in an international multi-centre study.

OBJECTIVE: This paper describes the composition and international multi-centre evaluation of a battery of tests termed the preliminary auditory profile. It includes measures of loudness perception, listening effort, speech perception, spectral and temporal resolution, spatial hearing, self-reported disability and handicap, and cognition. Clinical applicability and comparability across different centres are investigated. DESIGN: Headphone tests were conducted in five centres divided over four countries. Effects of test-retest, ear, and centre were investigated. Results for normally-hearing (NH) and hearing-impaired (HI) listeners are presented. STUDY SAMPLE: Thirty NH listeners aged 19-39 years, and 72 HI listeners aged 22-91 years with a broad range of hearing losses were included. RESULTS: Test-retest reliability was generally good and there were very few right/left ear effects. Results of all tests were comparable across centres for NH listeners after baseline correction to account for necessary differences between test materials. For HI listeners, results were comparable across centres for the language-independent tests. CONCLUSIONS: The auditory profile forms a clinical test battery that is applicable in four different languages. Even after baseline correction, differences between test materials have to be taken into account when interpreting results of language-dependent tests in HI listeners.

Measuring spectral and temporal resolution simultaneously: a comparison between two tests.

OBJECTIVE: Spectral- and temporal-resolution tests are seldom used in clinical practice despite their proven relevance for patients' speech understanding in noise and expected importance for hearing-aid fitting. The aim here was to investigate and compare two clinically applicable tests ('tone test' and 'sweep test') that measure both spectral and temporal resolution simultaneously. DESIGN: Experiments were conducted monaurally via headphones. After examining test-retest reliabilities and learning effects we compared results from tone and sweep tests to results from conventional spectral and temporal-resolution tests and to speech perception in noise scores. STUDY SAMPLE: A group of five normal-hearing listeners (aged 18-42 years, median 19) and 15 (sensorineurally) hearing-impaired listeners (aged 20-68 years, median 56). RESULTS: It was found that the tone test corresponded much better to the conventional methods than the sweep test. Relating spectral and temporal-resolution results to speech perception in noise scores showed that the tone test seems to be slightly more relevant for speech perception than the sweep test. CONCLUSIONS: It can be concluded that the tone test (after modifications we suggest, based on our findings) is a fast and reliable test that is suitable for measuring spectral and temporal resolution in a clinical setting.

Improved horizontal directional hearing in bone conduction device users with acquired unilateral conductive hearing loss.

We examined horizontal directional hearing in patients with acquired severe unilateral conductive hearing loss (UCHL). All patients (n = 12) had been fitted with a bone conduction device (BCD) to restore bilateral hearing. The patients were tested in the unaided (monaural) and aided (binaural) hearing condition. Five listeners without hearing loss were tested as a control group while listening with a monaural plug and earmuff, or with both ears (binaural). We randomly varied stimulus presentation levels to assess whether listeners relied on the acoustic head-shadow effect (HSE) for horizontal (azimuth) localization. Moreover, to prevent sound localization on the basis of monaural spectral shape cues from head and pinna, subjects were exposed to narrow band (1/3 octave) noises. We demonstrate that the BCD significantly improved sound localization in 8/12 of the UCHL patients. Interestingly, under monaural hearing (BCD off), we observed fairly good unaided azimuth localization performance in 4/12 of the patients. Our multiple regression analysis shows that all patients relied on the ambiguous HSE for localization. In contrast, acutely plugged control listeners did not employ the HSE. Our data confirm and further extend results of recent studies on the use of sound localization cues in chronic and acute monaural listening.