Lower Star Ratings Research Articles

Selecting Cover Images for Restaurant Reviews: AI vs. Wisdom of the Crowd

Problem definition: Restaurant review platforms, such as Yelp and TripAdvisor, routinely receive large numbers of photos in their review submissions. These photos provide significant value for users who seek to compare restaurants. In this context, the choice of cover images (i.e., representative photos of the restaurants) can greatly influence the level of user engagement on the platform. Unfortunately, selecting these images can be time consuming and often requires human intervention. At the same time, it is challenging to develop a systematic approach to assess the effectiveness of the selected images. Methodology/results: In this paper, we collaborate with a large review platform in Asia to investigate this problem. We discuss two image selection approaches, namely crowd-based and artificial intelligence (AI)-based systems. The AI-based system we use learns complex latent image features, which are further enhanced by transfer learning to overcome the scarcity of labeled data. We collaborate with the platform to deploy our AI-based system through a randomized field experiment to carefully compare both systems. We find that the AI-based system outperforms the crowd-based counterpart and boosts user engagement by 12.43%–16.05% on average. We then conduct empirical analyses on observational data to identify the underlying mechanisms that drive the superior performance of the AI-based system. Managerial implications: Finally, we infer from our findings that the AI-based system outperforms the crowd-based system for restaurants with (i) a longer tenure on the platform, (ii) a limited number of user-generated photos, (iii) a lower star rating, and (iv) lower user engagement during the crowd-based system. Funding: The authors acknowledge financial support from the Social Sciences and Humanities Research Council [Grant 430-2020-00106]. Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2021.0531 .

Classification of Patients' Judgments of Their Physicians in Web-Based Written Reviews Using Natural Language Processing: Algorithm Development and Validation.

Patients increasingly rely on web-based physician reviews to choose a physician and share their experiences. However, the unstructured text of these written reviews presents a challenge for researchers seeking to make inferences about patients' judgments. Methods previously used to identify patient judgments within reviews, such as hand-coding and dictionary-based approaches, have posed limitations to sample size and classification accuracy. Advanced natural language processing methods can help overcome these limitations and promote further analysis of physician reviews on these popular platforms. This study aims to train, test, and validate an advanced natural language processing algorithm for classifying the presence and valence of 2 dimensions of patient judgments in web-based physician reviews: interpersonal manner and technical competence. We sampled 345,053 reviews for 167,150 physicians across the United States from Healthgrades.com, a commercial web-based physician rating and review website. We hand-coded 2000 written reviews and used those reviews to train and test a transformer classification algorithm called the Robustly Optimized BERT (Bidirectional Encoder Representations from Transformers) Pretraining Approach (RoBERTa). The 2 fine-tuned models coded the reviews for the presence and positive or negative valence of patients' interpersonal manner or technical competence judgments of their physicians. We evaluated the performance of the 2 models against 200 hand-coded reviews and validated the models using the full sample of 345,053 RoBERTa-coded reviews. The interpersonal manner model was 90% accurate with precision of 0.89, recall of 0.90, and weighted F1-score of 0.89. The technical competence model was 90% accurate with precision of 0.91, recall of 0.90, and weighted F1-score of 0.90. Positive-valence judgments were associated with higher review star ratings whereas negative-valence judgments were associated with lower star ratings. Analysis of the data by review rating and physician gender corresponded with findings in prior literature. Our 2 classification models coded interpersonal manner and technical competence judgments with high precision, recall, and accuracy. These models were validated using review star ratings and results from previous research. RoBERTa can accurately classify unstructured, web-based review text at scale. Future work could explore the use of this algorithm with other textual data, such as social media posts and electronic health records.

Assessing Hospital Quality Scores By Proportion Of Patients Dually Eligible For Medicare And Medicaid.

The Centers for Medicare and Medicaid Services has been reporting hospital star ratings since 2016. Some stakeholders have criticized the star ratings methodology for not adjusting for social risk factors. We examined the relationship between 2021 star rating scores and hospitals' proportion of Medicare patients dually eligible for Medicaid. We found that, on average, hospitals caring for a greater proportion of dually eligible patients had lower star ratings, but there was significant overlap in performance among hospitals when we stratified them by quintile of dually eligible patients. Hospitals in the highest quintile (those with the greatest proportion of dually eligible patients) had the best mean mortality scores (0.28) but the worst readmission (-0.44) and patient experience (-0.78) scores. We assigned star ratings after stratifying the readmission measure group by proportion of dually eligible patients and found that a total of 142 hospitals gained a star and 161 hospitals lost a star, of which 126 (89percent) and 1 (<1percent) were in the highest quintile, respectively. Adjusting public reporting tools such as star ratings for social risk factors is ultimately a policy decision, and views on the appropriateness of accounting for factors such as proportion of dually eligible patients are mixed, depending on the organization and stakeholder.

Sentiment Analysis of Online Patient-Written Reviews of Vascular Surgeons

Online patient reviews influence a patient's choice of a vascular surgeon. The aim of this study is to examine underlying factors that contribute to positive and negative patient reviews by leveraging sentiment analysis and machine learning methods. The Society of Vascular Surgeons publicly accessible member directory was queried and cross-referenced with a popular patient-maintained physician review website, healthgrades.com. Sentiment analysis and machine learning methods were used to analyze several parameters. Demographics (gender, age, and state of practice), star rating (of 5 stars), and written reviews were obtained for corresponding vascular surgeons. A sentiment analysis model was applied to patient-written reviews and validated against the star ratings. Student's t-test or one-way analysis of variance assessed demographic relationships with reviews. Word frequency assessments and multivariable logistic regression analyses were conducted to identify common and determinative components of written reviews. A total of 1,799 vascular surgeons had public profiles with reviews. Female gender of surgeon was associated with lower star ratings (male=4.19, female=3.95, P<0.01) and average sentiment score (male=0.50, female=0.40, P<0.01). Younger physician age was associated with higher star rating (P=0.02) but not average sentiment score (P=0.12). In the Best reviews, the most commonly used one-words were Care (N=999), Caring (N=767), and Kind (N=479), while the most commonly used two-word pairs were Saved/Life (N=189), Feel/Comfortable (N=106), and Kind/Caring (N=104). For the Worst reviews, the most commonly used one-words were Pain (N=254) and Rude (N=148), while the most commonly used two-word pairs were No/One (N=27), Waste/Time (N=25), and Severe/Pain (N=18). In a multiple logistic regression, satisfactory reviews were associated with words such as Confident (odds ratio [OR]=8.93), Pain-free (OR=4.72), Listens (OR=2.55), and Bedside Manner (OR=1.70), while unsatisfactory reviews were associated with words such as Rude (OR=0.01), Arrogant (OR=0.09), Infection (OR=0.20), and Wait (OR=0.48). Female surgeons received significantly worse reviews and younger surgeons tended to receive better reviews. The positivity and negativity of reviews were largely related to words associated with the patient-doctor experience and pain. Vascular surgeons should focus on these 2 areas to improve patient experiences and their own reviews.

Higher Star, Better Performance in US Nursing Homes during the Covid-19 Pandemic ?

The Covid-19 pandemic has brought terrible difficulties to nursing homes, as they were locations with the highest number of confirmed Covid-19 cases and deaths in the US. The Centers for Medicare & Medicaid Services (CMS) applies the Five-Star Quality Ratings to indicate the quality of care in nursing homes based on health inspection survey, staffing, and resident outcome. Studies to date have contradictory findings regarding the relationship between nursing home reported Quality Ratings and Covid-19 cases and deaths based on US regional data. The purpose of this study is to examine whether nursing homes’ Quality Ratings were related to the total number of resident Covid-19 cases and deaths at the US National level. The study examined US nursing homes (N=13,494) linked with CMS Nursing Home Compare data and Covid-19 Nursing Home data. Using multiple linear regression analyses, results showed nursing home Quality Ratings were significantly associated with Covid-19 residents’ cases and deaths controlling for ownership type, size, occupation rate, and years of operation (p<.001; p<.001). Five-star nursing homes were less likely to have Covid-19 cases and deaths. Further, comparing lower Star Ratings nursing homes, 1-Star nursing homes showed no significant difference to 2-Star and 3-Star nursing homes when examining Covid-19 cases and deaths. Overall, the Five-Star Quality Ratings is a useful measure when investigating nursing homes’ performance during the Covid-19 pandemic. Future policymakers and administrators should also focus on nursing homes with lower star ratings when improving the quality of nursing homes, particularly with regard to resident health.

U.S. Nursing Home Quality Ratings Associated with COVID-19 Cases and Deaths

ObjectivesTo inform future policies and disaster preparedness plans in the vulnerable nursing home setting, we need greater insight into the relationship between nursing homes’ (NHs’) quality and the spread and severity of COVID-19 in NH facilities. We therefore extend current evidence on the relationships between NH quality and resident COVID-19 infection rates and deaths, taking into account NH structural characteristics and community characteristics. DesignCross-sectional study. Setting and Participants15,390 Medicaid- and Medicare-certified NHs. MethodsWe obtained and merged the following data sets: (1) COVID-19 weekly data reported by each nursing home to the Centers for Disease Control and Prevention’s National Healthcare Safety Network, (2) Centers for Medicare & Medicaid Services Five Star Quality Rating System, (3) county-level COVID-19 case counts, (4) county-level population data, and (5) county-level sociodemographic data. ResultsAmong 1-star NHs, there were an average of 13.19 cases and 2.42 deaths per 1000 residents per week between May 25 and December 20, 2020. Among 5-star NHs, there were an average of 9.99 cases and 1.83 deaths per 1000 residents per week. The rate of confirmed cases of COVID-19 was 31% higher among 1-star NHs compared with 5-star NHs [model 1: incidence rate ratio (IRR) 1.31, 95% confidence interval (CI) 1.23-1.39], and the rate of COVID-19 deaths was 30% higher (IRR 1.30, 95% CI 1.20, 1.41). These associations were only partially explained by differences in community spread of COVID-19, case mix, and the for-profit status and size of NHs. Conclusions and ImplicationsWe found that COVID-19 case and death rates were substantially higher among NHs with lower star ratings, suggesting that NHs with quality much below average are more susceptible to the spread of COVID-19. This relationship, particularly with regard to case rates, can be partially attributed to external factors: lower-rated NHs are often located in areas with greater COVID-19 community spread and serve more socioeconomically vulnerable residents than higher-rated NHs.

Selecting Cover Images for Restaurant Reviews: AI vs. Wisdom of the Crowd

Restaurant review platforms, such as Yelp and Tripadvisor, routinely receive large numbers of photos in their review submissions. These photos provide significant value for users who seek to compare restaurants. In this context, the choice of cover images (i.e., representative photos of the restaurants) can greatly influence the level of user engagement on the platform. Unfortunately, selecting these images can be time consuming and often requires human intervention. At the same time, it is challenging to develop a systematic approach to assess the effectiveness of the selected images. In this paper, we collaborate with a large review platform in Asia to investigate this problem. We discuss two image selection approaches, namely crowd-based and AI-based systems. The AI-based system we propose learn complex latent image features, which is further enhanced by transfer learning to overcome the scarcity of labeled data. We collaborate with the platform to deploy our AI-based system through a randomized field experiment to carefully compare both systems. We find that the AI-based system outperforms the crowd-based counterpart and boosts user engagement by 12.43%-16.05% on average. We then conduct empirical analyses on observational data to identify the underlying mechanisms that drive the superior performance of the AI-based system. Finally, we infer from our findings that the AI-based system outperforms the crowd-based system for restaurants with a (i) longer tenure on the platform, (ii) more expensive offering, (iii) lower star rating, and (iv) limited number of user-generated photos.

Quality of Post-Acute Care in Skilled Nursing Facilities That Disproportionately Serve Hispanics With Dementia

ObjectivesAs the number of Hispanics with dementia continues to increase, greater use of post-acute care in nursing home settings will be required. Little is known about the quality of skilled nursing facilities (SNFs) that disproportionately serve Hispanic patients with dementia and whether the quality of SNF care varies by the concentration of Medicare Advantage (MA) patients with dementia admitted to these SNFs. DesignCross-sectional study using 2016 data from Medicare certified providers. Setting and ParticipantsOur cohort included 177,396 beneficiaries with probable dementia from 8884 SNFs. MethodsWe examined facility-level quality of care among facilities with high and low proportions of Hispanic beneficiaries with probable dementia enrolled in MA and fee-for-service (FFS) using data from Medicare-certified providers. Three facility-level measures were used to assess quality of care: (1) 30-day rehospitalization rate; (2) successful discharge from the facility to the community; and (3) Medicare 5-star quality ratings. ResultsAbout 20% of residents were admitted to 1615 facilities with a resident population that was more than 15% Hispanic. Facilities with a higher share of Hispanic residents had a lower proportion of 4- or 5-star facilities by an average of 14% to 15% compared with facilities with little to no Hispanics. In addition, these facilities had a 1% higher readmission rate. There were also some differences in the quality of facilities with high (>26.5%) and low (<26.5%) proportions of MA beneficiaries. On average, SNFs with a high concentration of MA patients have lower readmission rates and higher successful discharge, but lower star ratings. Conclusions and ImplicationsAchieving better quality of care for people with dementia may require efforts to improve the quality of care among facilities with a high concentration of Hispanic residents.

HIGH ON SMI: QUALITY CONCERNS ABOUT NURSING HOMES SERVING HIGH PROPORTIONS OF SERIOUSLY MENTALLY ILL RESIDENTS

The proportion of nursing home (NH) residents that have serious mental illness (SMI) has increased over the least two decades. Residents with SMI tend be younger and have different medical needs than traditional residents. To better understand this population, our study examined the facility, staffing, and resident characteristics of NHs that were more likely to specialize in SMI. Utilizing the Certification and Survey Provider Enhanced Reports, low-SMI (N = 3,616) and high-SMI (N = 3,615) NHs were defined as the first and fourth quartile of the distribution of the proportion of SMI residents, respectively. We performed bivariate tests and multivariate logistic regression to compare facility, staffing, resident, and star-ratings characteristics between NHs. High-SMI NHs were less likely to be Not-For-Profit, have fewer beds, have more Medicaid-paying residents, lower registered nurse staffing, and lower certified nurse aide staffing levels (p’s<.001). Residents in high-SMI NHs were more likely to require behavioral healthcare (p<.001) and be treated with psychoactive medications (any psychoactive, antidepressants, antipsychotics, anxiolytics (p’s<.001), hypnotics (p<.01)). Finally, high-SMI facilities had lower overall quality, health inspection, quality measure, staffing, and registered nurse staffing star-ratings (p’s<.001). High-SMI NHs have characteristics that are associated with lower quality-of-care (e.g., For-Profit, more Medicaid), lower staffing, prescribe more psychoactive medications, and have lower star-ratings. As the SMI population grows, large numbers of SMI residents will concentrate in a few NHs. While further research is needed to understand the implications of these trends, policy-makers must be aware of this population when affecting the resources and staffing of NHs.

Same sushi, different impressions: a cross-cultural analysis of Yelp reviews

Online reviews are essential digital assets for any service business in the hospitality industry. This study analyzes 76,704 Western (North American and European) and 56,159 Japanese Yelp reviews of Japanese restaurants. We find that Western consumers are more likely to give higher or lower star ratings than are Japanese consumers, while Japanese consumers are more likely to vote on the helpfulness of others’ reviews than are Westerners. Further analyses of the review texts show that Western and Japanese consumers express their sentiments over different dimensions of restaurant experience (food quality, service quality, the physical environment, and price fairness) for the same categories of Japanese dish. Westerners express more sentiments on service quality overall; Japanese customers express more negative sentiments on inferior physical environments and more positive sentiments on price fairness. These findings indicate that culture influences the consumer experience expressed in online reviews. Our study thus offers insights into online travel portals and restaurant practitioners to help them use online reviews wisely and appropriately accommodate customers with varied cultural backgrounds.

Star power: the effect of star rating on service recovery in the hotel industry

PurposeThis study investigated distributive justice (DJ), procedural justice (PJ) and interactional justice (IJ) in the recovery of service in the hotel industry as well as their effects on post-recovery satisfaction and the subsequent behavior of customers. This study also sought to elucidate the moderating effects of star ratings on these influences.Design/methodology/approachA field experiment design with 424 consumers using structural equation modeling was adopted to test the study hypotheses.FindingsDJ, PJ and IJ were all shown to have a significantly positive effect on post-recovery satisfaction, which in turn has a significantly positive effect on customer loyalty and significantly negative effect on customer defection. Analysis on the moderating effects of star ratings also revealed that the relationship between DJ and post-recovery satisfaction during recovery efforts is stronger for hotels with higher star ratings than for hotels with lower star ratings. During recovery, post-recovery satisfaction was shown to be more effective in reducing customer defection from hotels with lower star ratings than from hotels with higher star ratings.Practical implicationsThis study provides a reference by which hotels can match their service recovery mechanism to their star rating to retain customers and increase customer loyalty in the event of service failure.Originality/valueThis study confirms the importance of IJ in the service recovery processes in the hotel industry and provides verification that the star rating of a hotel plays a crucial role as a moderator between recovery effort and recovery effectiveness.

Medicare Advantage Ratings And Voluntary Disenrollment Among Patients With End-Stage Renal Disease.

Populations with intensive health care needs and high care costs may be attracted to insurance plans that have high quality ratings, but patients may be likely to disenroll from a plan if their care needs are not met. We assessed the association between publicly reported Medicare Advantage plan star ratings and voluntary disenrollment of incident dialysis patients in the following year over the period 2007-13. We found that Medicare Advantage (MA) plans with lower star ratings had significantly higher rates of disenrollment by incident dialysis patients in the following year. Compared to MA plans with 4.0 or more stars, adjusted disenrollment rates were 3.9 percentage points higher for plans with 3.5 stars, 5.0 percentage points higher for those with 3.0 stars, and 12.1 percentage points higher for those with 2.5 or fewer stars. These findings suggest that low plan quality may lead to increased expenditures, as this high-cost population generally must shift from Medicare Advantage to traditional Medicare upon disenrollment.

A Modified Model for Hotel Website Functionality Evaluation

ABSTRACTHarnessing a mathematical computation approach to analyze data solicited from 354 website stakeholders (including 200 hotel customers, 83 hotel managers, and 71 hotel website designers), this study presents an up-to-date and robust evaluation model for assessing the functionality of hotel websites. The modified model was also empirically used to assess the functionality performance of the websites of all the hotels in Hong Kong. Results from the non-parametric Kruskal–Wallis and Mann–Whitney tests revealed a difference in performance, with higher star rating hotels performing better than those with a lower star rating. As one of the first attempts to incorporate the perceptions of multiple website stakeholders into hotel website evaluation, this study provides hoteliers with practical advice on website development.