Social Network Analysis of COVID-19 Sentiments: Application of Artificial Intelligence

Background

The coronavirus disease (COVID-19) pandemic led to substantial public discussion. Understanding these discussions can help institutions, governments, and individuals navigate the pandemic.

Objective

The aim of this study is to analyze discussions on Twitter related to COVID-19 and to investigate the sentiments toward COVID-19.

Methods

This study applied machine learning methods in the field of artificial intelligence to analyze data collected from Twitter. Using tweets originating exclusively in the United States and written in English during the 1-month period from March 20 to April 19, 2020, the study examined COVID-19–related discussions. Social network and sentiment analyses were also conducted to determine the social network of dominant topics and whether the tweets expressed positive, neutral, or negative sentiments. Geographic analysis of the tweets was also conducted.

Results

There were a total of 14,180,603 likes, 863,411 replies, 3,087,812 retweets, and 641,381 mentions in tweets during the study timeframe. Out of 902,138 tweets analyzed, sentiment analysis classified 434,254 (48.2%) tweets as having a positive sentiment, 187,042 (20.7%) as neutral, and 280,842 (31.1%) as negative. The study identified 5 dominant themes among COVID-19–related tweets: health care environment, emotional support, business economy, social change, and psychological stress. Alaska, Wyoming, New Mexico, Pennsylvania, and Florida were the states expressing the most negative sentiment while Vermont, North Dakota, Utah, Colorado, Tennessee, and North Carolina conveyed the most positive sentiment.

Conclusions

This study identified 5 prevalent themes of COVID-19 discussion with sentiments ranging from positive to negative. These themes and sentiments can clarify the public’s response to COVID-19 and help officials navigate the pandemic.

A Normative Dataset of the Balance Error Scoring System in Children Aged Between 5 and 14

OBJECTIVE

Pediatric head injuries occur commonly and are being reported in increasing numbers. Balance testing is a key component in the evaluation of suspected concussion, and the balance error scoring system (BESS) is likely the most well-known and widely used measure. To date, normative BESS scores for adults have been reported but not for children.

DESIGN

Normative data for BESS scores and modified BESS scores were created in a cohort of healthy children. Potential variables were analyzed as predictors of BESS performance.

SETTING

Local elementary and junior high schools.

PARTICIPANTS

A total of 373 healthy children between the ages of 5 and 14.

INTERVENTIONS

The BESS was performed on all children.

ASSESSMENT OF RISK FACTORS

Gender, body mass index percentile, previous concussions, athletic participation, age, and the parental opinion of child's balance ability were examined as factors associated with the BESS score.

MAIN OUTCOME MEASURES

BESS scores.

RESULTS

Normative data are reported, stratified by age groups of 5 to 7 years, 8 to 10 years, and 11 to 14 years of age, for both BESS and modified BESS. Median BESS scores are 23 for children aged 5 to 7, 18 for children aged 8 to 10, and 16 for children aged 11 to 14. Median modified BESS scores are 8 for children age 5 to 7, 5 for children age 8 to 10, and 4 for children age 11 to 14. Increasing age and positive parental opinion regarding their child's balance ability were independently correlated with decreasing BESS scores (P < 0.01).

CONCLUSIONS

The normative data on BESS scores for healthy children reported here provide age-stratified reference values for suspected balance alterations.

Challenging the norm: further psychometric investigation of the neck disability index

BACKGROUND CONTEXT

The neck disability index (NDI) was the first patient-reported outcome (PRO) instrument specific to patients with neck pain, and it remains one of the most widely used PROs for the neck population. The NDI is an appealing measure as it is a short and well-known PRO measure. Currently, there are conflicting data on the performance and applicability of the NDI in patients undergoing either operative or nonoperative treatment for neck-related conditions.

PURPOSE

This study investigates the psychometric properties, performance, and applicability of the NDI in the spine patient population.

STUDY DESIGN

A total of 865 patients visiting a university-based spine clinic with neck complaints, with or without radiating upper extremity pain, numbness, or weakness were enrolled in the study. Visit types included new and follow-up visits to both operative and nonoperative treatments. Questionnaires were administered electronically on a tablet computer, and all patients answered all 10 questions of the NDI.

METHODS

Standard descriptive statistics were performed to describe the demographic characteristics of the patients. Rasch modeling was applied to examine the psychometric properties of the NDI.

RESULTS

The NDI demonstrated insufficient unidimensionality (ie, unexplained variance after accounting for the first dimension=9.4%). Person reliability was 0.85 and item reliability was 1.00 for the NDI. The overall item fit for the NDI was good with an outfit mean square of 1.03. The NDI had a floor effect of 35.5% and ceiling effect of 4.6%. The raw score to measure correlation of the NDI was 0.019.

CONCLUSIONS

Although the NDI had good person and item reliability, it did not demonstrate strong evidence of unidimensionality. The NDI exhibited a very large floor effect. Because of the poor raw score to measure correlation, the sum score should not be used in interpretation of findings. Despite great investment by physicians and other stakeholders in the NDI, this evaluation and previous research have demonstrated that the NDI needs further investigation and refinement.

Patient-Reported Outcomes and Total Health Care Expenditure in Prediction of Patient Satisfaction: Results From a National Study

Background

Health care quality is often linked to patient satisfaction. Yet, there is a lack of national studies examining the relationship between patient satisfaction, patient-reported outcomes, and medical expenditure.

Objective

The aim of this study is to examine the contribution of physical health, mental health, general health, and total health care expenditures to patient satisfaction using a longitudinal, nationally representative sample.

Methods

Using data from the 2010-2011 Medical Expenditure Panel Survey, analyses were conducted to predict patient satisfaction from patient-reported outcomes and total health care expenditures. The study sample consisted of adult participants (N=10,157), with sampling weights representative of 233.26 million people in the United States.

Results

The results indicated that patient-reported outcomes and total health care expenditure were associated with patient satisfaction such that higher physical and mental function, higher general health status, and higher total health care expenditure were associated with higher patient satisfaction.

Conclusions

We found that patient-reported outcomes and total health care expenditure had a significant relationship with patient satisfaction. As more emphasis is placed on health care value and quality, this area of research will become increasingly needed and critical questions should be asked about what we value in health care and whether we can find a balance between patient satisfaction, outcomes, and expenditures. Future research should apply big data analytics to investigate whether there is a differential effect of patient-reported outcomes and medical expenditures on patient satisfaction across different medical specialties.

Effectiveness of the extended parallel process model in promoting colorectal cancer screening

OBJECTIVE

Relatives of colorectal cancer (CRC) patients are at increased risk for the disease, yet screening rates still remain low. Guided by the Extended Parallel Process Model, we examined the impact of a personalized, remote risk communication intervention on behavioral intention and colonoscopy uptake in relatives of CRC patients, assessing the original additive model and an alternative model in which each theoretical construct contributes uniquely.

METHODS

We collected intention-to-screen and medical record-verified colonoscopy information on 218 individuals who received the personalized intervention.

RESULTS

Structural equation modeling showed poor main model fit (root mean square error of approximation (RMSEA) = 0.109; standardized root mean residual (SRMR) = 0.134; comparative fit index (CFI) = 0.797; Akaike information criterion (AIC) = 11,601; Bayesian information criterion (BIC) = 11,884). However, the alternative model (RMSEA = 0.070; SRMR = 0.105; CFI = 0.918; AIC = 11,186; BIC = 11,498) showed good fit. Cancer susceptibility (B = 0.319, p < 0.001) and colonoscopy self-efficacy (B = 0.364, p < 0.001) perceptions predicted intention to screen, which was significantly associated with colonoscopy uptake (B = 0.539, p < 0.001).

CONCLUSIONS

Our findings provide support of the utility of Extended Parallel Process Model for designing effective interventions to motivate CRC screening in persons at increased risk when individual elements of the model are considered. Copyright © 2015 John Wiley & Sons, Ltd.

Intention to undergo colonoscopy screening among relatives of colorectal cancer cases: a theory-based model

BACKGROUND

It is recommended that persons having familial risk of colorectal cancer begin regular colonoscopy screening at an earlier age than those in the general population. However, many individuals at increased risk do not adhere to these screening recommendations.

PURPOSE

The goal of this study was to examine cognitive, affective, social, and behavioral motivators of colonoscopy intention among individuals at increased risk of familial colorectal cancer.

METHODS

Relatives of colorectal cancer cases (N = 481) eligible for colonoscopy screening completed a survey assessing constructs from several theoretical frameworks including fear appeal theories.

RESULTS

Structural equation modeling indicated that perceived colorectal cancer risk, past colonoscopy, fear of colorectal cancer, support from family and friends, and health-care provider recommendation were determinants of colonoscopy intention.

CONCLUSIONS

Future interventions to promote colonoscopy in this increased risk population should target the factors we identified as motivators. (ClinicalTrials.gov number NCT01274143).

Psychometric Evaluation of the Lower Extremity Computerized Adaptive Test, the Modified Harris Hip Score, and the Hip Outcome Score

Background:

The applicability and validity of many patient-reported outcome measures in the high-functioning population are not well understood.

Purpose:

To compare the psychometric properties of the modified Harris Hip Score (mHHS), the Hip Outcome Score activities of daily living subscale (HOS-ADL) and sports (HOS-sports), and the Lower Extremity Computerized Adaptive Test (LE CAT). The hypotheses was that all instruments would perform well but that the LE CAT would show superiority psychometrically because a combination of CAT and a large item bank allows for a high degree of measurement precision.

Study Design:

Cohort study (diagnosis); Level of evidence, 2.

Methods:

Data were collected from 472 advanced-age, active participants from the Huntsman World Senior Games in 2012. Validity evidences were examined through item fit, dimensionality, monotonicity, local independence, differential item functioning, person raw score to measure correlation, and instrument coverage (ie, ceiling and floor effects), and reliability evidences were examined through Cronbach alpha and person separation index.

Results:

All instruments demonstrated good item fit, unidimensionality, monotonicity, local independence, and person raw score to measure correlations. The HOS-ADL had high ceiling effects of 36.02%, and the mHHS had ceiling effects of 27.54%. The LE CAT had ceiling effects of 8.47%, and the HOS-sports had no ceiling effects. None of the instruments had any floor effects. The mHHS had a very low Cronbach alpha of 0.41 and an extremely low person separation index of 0.08. Reliabilities for the LE CAT were excellent and for the HOS-ADL and HOS-sports were good.

Conclusion:

The LE CAT showed better psychometric properties overall than the HOS-ADL, HOS-sports, and mHHS for the senior population. The mHHS demonstrated pronounced ceiling effects and poor reliabilities that should be of concern. The high ceiling effects for the HOS-ADL were also of concern. The LE CAT was superior in all psychometric aspects examined in this study. Future research should investigate the LE CAT for wider use in different populations.

Evaluation of the PROMIS physical function computer adaptive test in the upper extremity

PURPOSE

To compare psychometric and responder burden characteristics between the Patient-Reported Outcomes Measurement Information System Physical Function Computer Adaptive Test (PF CAT) and the Disabilities of Arm, Shoulder, and Hand (DASH) instruments in a tertiary hand and upper extremity practice.

METHODS

Adult patients who presented to the clinic of 2 hand and upper extremity surgeons in a university-based tertiary care center were enrolled in this study. Participants received the DASH and PF CAT administered via tablet computer. Time to completion was recorded for both the DASH and PF CAT. We conducted statistical analyses to calculate Pearson correlation coefficients between the 2 instruments and performed a Rasch item response theory analysis to determine dimensionality, reliability, ceiling and floor effects, and item bias for each instrument.

RESULTS

A total of 134 patients were included. Time to completion for the DASH was 262 seconds, and for the PF CAT 57 seconds. The instruments had strong correlation (r = 0.726). The item and Pearson reliability were 0.97 and 0.94, respectively, for the DASH and 0.99 and 0.96 for the PF CAT. The DASH and PF CAT had 5% and 5% of unexplained variance, respectively. The DASH exhibited 5% of ceiling effect and 1% floor effect whereas the PF CAT had no ceiling or floor effects.

CONCLUSIONS

The psychometric characteristics of the Patient-Reported Outcomes Measurement Information System PF CAT instrument compared favorably with the DASH in a tertiary upper extremity practice. Patient time burden was significantly reduced with the PF CAT compared with the DASH.

TYPE OF STUDY/LEVEL OF EVIDENCE

Diagnostic III.

Psychometric Comparison of the PROMIS Physical Function CAT With the FAAM and FFI for Measuring Patient-Reported Outcomes

BACKGROUND

Selecting optimal patient-reported outcome (PRO) instruments is critical to improving the quality of health care. The purpose of this study was to compare the reliability, responsiveness, and efficiency of three PRO measures: the Foot and Ankle Ability Measure-Activity of Daily Living subscale (FAAM_ADL), the Foot Function Index 5-point verbal rating scale (FFI-5pt), and the PROMIS Physical Function computerized adaptive test (PF CAT).

METHODS

Data were aggregated from 10 clinical sites in the AOFAS's National Orthopaedic Foot and Ankle Research (OFAR) Network from 311 patients who underwent elective surgery for a disorder of the foot or ankle. Patients were administered the FAAM_ADL, FFI-5pt, and PF CAT at their preoperative visit and at 6 months after surgery. Reliabilities were evaluated using a Rasch model. Responsiveness was calculated using paired samples t test and efficiency was recorded as number of seconds to complete the instrument.

RESULTS

Similar reliabilities were found for the three instruments. Item reliabilities for FAAM_ADL, FFI-5pt, and PF CAT were all .99. Pearson reliabilities for FAAM_ADL, FFI-5pt, and PF CAT were .95, .93, and .96, respectively. On average, patients completed the FAAM_ADL in 179 seconds, the FFI-5pt in 194 seconds, and the PF CAT in 44 seconds, ( P < .001). The PF CAT and FAAM_ADL showed significant improvement ( P = .01 and P = .001, respectively) in patients' physical function after treatment; the FFI-5pt did not show improvement.

CONCLUSIONS

Overall, the PF CAT performed best in terms of reliability, responsiveness, and efficiency in this broad sample of foot and ankle patients. It can be a potential replacement for the conventional PRO measures, but further validation is needed in conjunction with the PROMIS Pain instruments.

LEVEL OF EVIDENCE

Level I, prospective comparative outcome study.

Time for a paradigm shift with computerized adaptive testing of general physical function outcomes measurements

BACKGROUND

Patient-reported outcomes (PRO) are critical to understanding the value of orthopedic treatments. We hypothesized that use of the computerized adaptive testing from a well-characterized physical function item bank would show superiority in assessing all levels of physical function compared to current standard generic physical function outcomes instruments for foot and ankle patients.

METHODS

In a population of 126 foot and ankle patients we compared the psychometric properties for three PROs: the Physical Function subscale of the Medical Outcomes General Health Survey (SF-36 PF) version 2, the Patient-Reported Outcomes Measurement Information System (PROMIS) Physical Function (PF) Computerized Adaptive Test (CAT), and the Lower Extremity (LE) CAT. A Rasch item response theory (IRT) model was applied to assess and compare the fit, dimensionality, reliability, validity, and coverage.

RESULTS

The unexplained variance for the PF CAT was 3.9% and the LE CAT was 2.1%, suggesting each instrument explained a single concept. The SF-36 PF had more concerning unexplained variance of 7.6%. We found no floor or ceiling effects for the PF CAT, a minimal floor effect (1.6%) but no ceiling effect for the LE CAT, and an 11.1% floor effect and 9.5% ceiling effect for the SF-36 PF.

CONCLUSION

Foot and ankle clinicians and researchers interested in measuring patient perceived functional outcomes with a generic instrument should consider using either the PF CAT or the LE CAT rather than the SF-36 PF. Further studies comparing these CATs to anatomic specific instruments are needed.

LEVEL OF EVIDENCE

Level I, diagnostic study.

Uncovering patterns of technology use in consumer health informatics

Internet usage and accessibility has grown at a staggering rate, influencing technology use for healthcare purposes. The amount of health information technology (Health IT) available through the Internet is immeasurable and growing daily. Health IT is now seen as a fundamental aspect of patient care as it stimulates patient engagement and encourages personal health management. It is increasingly important to understand consumer health IT patterns including who is using specific technologies, how technologies are accessed, factors associated with use, and perceived benefits. To fully uncover consumer patterns it is imperative to recognize common barriers and which groups they disproportionately affect. Finally, exploring future demand and predictions will expose significant opportunities for health IT. The most frequently used health information technologies by consumers are gathering information online, mobile health (mHealth) technologies, and personal health records (PHRs). Gathering health information online is the favored pathway for healthcare consumers as it is used by more consumers and more frequently than any other technology. In regard to mHealth technologies, minority Americans, compared with White Americans utilize social media, mobile Internet, and mobile applications more frequently. Consumers believe PHRs are the most beneficial health IT. PHR usage is increasing rapidly due to PHR integration with provider health systems and health insurance plans. Key issues that have to be explicitly addressed in health IT are privacy and security concerns, health literacy, unawareness, and usability. Privacy and security concerns are rated the number one reason for the slow rate of health IT adoption.