Kanak RoyChowdhury

One of the biggest challenges of healthcare industry is frequent influx of new joiners, who are fresh from institutions, having lack of professional exposure, unclear about the application of the SOPs and unable to convince the customers about their queries due to the non confident response. In order to overcome these issues long learning curve seems to be most appropriate which isn't feasible in general. In order to curb this, AI can help in many ways. 1. Agent Assist BOT: basis the performance of agent, real time feedback can be generated 2. Training and Assessment Module: AI can create a platform, where multiple simulation can be generated which will be handled by agents and their responses will be assessed and feedback will be shared. It would help them to acquire real life challenges 3. Chat BOTs: learning can be made autonomous with the help of bot, which new joiners can use as per their requirement 24x7. They can create imaginary scenarios and handle them to gain confidence 4. Audit feedback is another method to educate the new joiners which is limited due to sample audit. Using automated audits larger sample size hence the feedback can be generated and circulated substantially. 5. Recruitment of appropriate candidate: AI can ensure fare and emotion less decision making which will help recruitment of appropriate candidates basis the requirement. This will help to prevent attrition & termination of the non fitting candidates

AI based solutions are disruptive in nature & can bring many types of improvements across the industries. These can be used to improve PQCDSM through systematic implementation & selection of appropriate use cases. Healthcare industries being one of the most organized domains are suitable for deployment of various AI aspects because of their regulations, customer orientation, easy availability of end-to-end information & many more. These breakthrough ideas can be deployed from the beginning of the value chain & can be applicable upto the end point. Examples are discussed in detail below: 1. Contracts or SOW Assessment: Details of the agreement are lengthy & require long time to interpret & collaborate. LLM based BOTs can be used to summarize the contract details 2. Call Deflection: Using real time AI based solution, calls from patients/services providers can be deflected to appropriate channels (voice/chat) where waiting time will be minimum 3. Chatbots & Agent assist: AI based chatbots can be deployed to answer simpler queries to improve profitability & round the clock availability. Agent assistance is another tool which can be used to simplify agents’ research activity & improve accuracy & productivity. 4. Accent Neutralization & Translation: AI based accent translator can be deployed to neutralize regional accent effect. Translators also can be used to address wide range of diverse speaking customers 5. Realtime Volume Forecasting: AI based solution can be used to do real time volume forecasting analyzing huge data & forecast with lowest MAPE. This will help to optimize manpower 6. Sentiment Analysis: Realtime call analysis will help to do sentiment analysis of the callers & basis the sentiment of the callers required corrective measures can be taken even by an inexperienced agent 7. Quality Audits: AI based solution can be used to audit larger samples & take proactive corrective actions to reduce FCR (First Call Resolution) related defects 8. Conversational Agents: Advocates can use it to convert text to speech or vice versa which will reduce their inefficiency in any particular mode of conversation 9. Fraud Detection: ML can help to identify anomalies from the submitted claims, which will help advocates to reduce research 10. Documentation: Mandatory documentation after each conversation can be automated using AI, which will reduce engage time of the advocates & will increase their availability for fresh conversations 11. Centralized Reporting System: Smart real time reporting system will help customers to understand processing status of transactions independently For deployment of these ideas a mix of top down & bottom up approach to be taken up by the organizations. So that, expertise of every level right from business leaders, AI developers & AI practitioners can be gathered to get the the fullest advantage.

Working of AI can be broadly categorized into 4 types: 1) Creation 2) Summarization 3) Discover & 4) Automate. Discover capability can be leveraged for the following: identification of hidden patterns & insights from data, searching of resources & docs, monitoring of real time events. Models can be trained on data, which are captured from server or sensors for analyzing & predicting future events like performance trend, CSAT, quality defects, temperature rise or machine breakdown etc. Similarly using ML, unlabeled historical data can be analyzed to identify unusual pattern of the data. Contract maintenance of existing service providers or fresh contracting of new service providers are considered the brain of healthcare industry. This eventually governs everything in healthcare business. Any mistake if remains in the contract, both auto and manual claim adjudication will get impacted in terms of payment of service providers, despatch of cheque to the correct address, service providers may not get payment as per revised rates or for the correct diagnosis. At the same time patients may not be able to book for appointments as per their requirement if contact details, language, visit hours are not updated. They can get delayed service from facilities if payment issue persists. These will generate D-SAT among facilities, doctors and patients. Impact of these errors are far reaching for insurance companies. To mitigate these errors, multiple rework queues, different layers of quality audit, calls/email/chat to customer care center are required which increase claim processing cost by manifold which decreases profitability of the insurance companies. AI can be used effectively to handle this challenge. Leveraging summarisation capability of Gen AI be helpful for decoding large contract details. In case of missing information automated alert can be raised to contract reviewing authority or network managers. Dependencies on individual processors' capability will be completely removed for identifying misses and raising clarification. Along with misses if any abnormal information is there that also can be flagged and can be double checked by deploying HITL. Once the clarifications are received, notification to the respective processors with TAT can be generated so that they can prioritize amid of their regular jobs accordingly. History of these network managers or organisation can be further patterned and contact renewal success probability can be displayed to them during their online annual renewal. Many times contract renewal comes with Special loading instructions along with regular contract documents. These special loading instructions are free text based documents and time consuming for processing. Using ML, processing of these free texts can be automated by updating the database as per instructions or clarifications can be raised. Therefore, using AI contract data management can be further simplified, job satisfaction, accuracy and C-SAT of the stakeholders can be increased, profitability of the insurance companies can be increased by reducing a huge amount of rework, processing time, call reduction etc.

In BPO industry, WFM plays a crucial role by planning resources in critical processing areas. This is achieved by factoring many Xs, which are experienced over time, learned over mistakes & so on. Therefore, any changes to the developed model is done after many iteration because of lack of visibility in manual calculation. One of the major factors behind AI based workforce scheduling is consideration or inclusion of real time variations e.g. demand variation, shrinkage, required run rate of SLAs, handling of specialized transactions, allocation of skilled agents for specific transactions, downtime of any app or platform etc. As because all of these calculations can be coupled with company policies & local labour laws, fairness to employee will be automatically ensured & not be biased or depended upon any human being. For achieving accuracy, different time series forecasting methods e.g. winter holt or other exponential smoothening techniques can be coupled & output with lowest MAPE results will only be suggested. Beside these, HITL to be considered to reduce excessive dependency on the data or biasness.

With advent of time, business has become more knowledge based. Detailed documentation of the process, flow charts, , drawings, standards, decision trees, QAPs, legal compliances, periodic updates of SOPs have become integral part of efficient operations. Processors/operators need to refer these document every now & then to main required CSAT, targeted revenue, escalations etc. Conventional ways of referring may not be effective when you are quoting a benefit of medical policy to an online caller because it would be time consuming as multiple clicks through contract details would take the agents to desired clauses. If a LLM based model coupled with Semantic technology is enabled here, the agent can use conversational AI to arrive quickly. Chatbots can be used to refer previous similar questions or advanced FAQ models. Similarly decoding of a complex claim history would become easier if SOPs, remark codes, diagnostic codes are connected using Ontario technology. This will not only make the process efficient & simpler but dependency on a SME will become negligible. Learning curve of average agents will get reduced significantly.

Conventional decision-making process is like a structured problem-solving process, which involves understanding problem statements, data collection, analyzing scenarios & arrive at the conclusion. Similarly, AI models analyze given set of data, identify patterns, train themselves and calculate the likelihood of different outcomes for both old & new cases. Therefore, case-based AI models can be developed for decision making for administrative as well as for functional scenario. The models may have different limitations like biases, hallucinations etc. which can be handled using techniques like Prompt engineering – develop precise prompts to guide the model, fine-tuning or further training, grounding of models using specific data & RAG – feedback mechanism. Success of any model depends upon its accuracy and accuracy of any AI model depends upon the following: 1) Data Accuracy 2) Enough data to make an accurate prediction 3) inclusive & representative data to avoid biased prediction 4) In scope data perfectly aligned with intended model 5) Consistent & labeled data. However for high risk & sensitive decision making models must be accompanied with HITL to ensure appropriate content creation with post generation review

Dear Sir, Healthcare VSM is not available here...

Multicollinearity is a phenomenon that happens during multiple regression analysis. In multiple regression analysis, the dependent variable (Y) is dependent on nos. of independent variables (X1, X2, X3 etc.). Every independent variable has their own influence on Y separately & the regression equation will reflect their individual effect in a combined manner in the form of Y=a+bX1+cX2+dX3. These independent variables are not dependent on each other. Therefore, X1, X2, X3 etc will influence only Y & not themselves. This equation implies that, for a “b” unit change of Y, X1 changes by 1 & X2 & X3 will remain constant. Similarly, when X2 changes by 1 unit, Y changes by “c” & X1 & X3 will remain constant & so on. On the other hand, if variables are selected in such a manner that, they are related to each other & change in 1 independent variable will change other independent variable instead of remaining constant, it will create problems on interpretation of behavior of individual independent variables. This phenomenon is known as multicollinearity. If multicolinearity occurs following problem may occur: i) Instead of having high R², any independent variable (s) may be non-significant (generally>0.05), means the variable doesn’t have any relation with Y ii) Sign of the coefficient of the independent variable may differ from real life understanding iii) Standard error will be large Detection: i) From the above 3 signs, multicollinearity can be detected ii) From the VIF (variable inflation factor) value, which is equal to 1/(1- R²). If VIF>5; high multicollinearity. If 1<VIF<3, moderate multicollinearity & can be ignored iii) From the r values between each two independent variables, if r values are too high; close to 1 that means multicollinearity may occur Course of action when detected: i) Conduct regression analysis considering one of the independent variables as Y and rest as Xs in a rotational manner & calculate VIF as mentioned above for each regression ii) Identify the independent variable (which is in the position of Y) with high VIF (Minitab analysis directly shows high VIF) iii) Remove the independent variable with high VIF & conduct the regression once again iv) In Minitab, stepwise regression analysis can be choosen with α=0.05 to remove the variable directly iv)If F is significant then overall model will be significant & the equation can be considered if effect of individual coefficients not to be analysed. It is better to remove the correlated variables & conduct the analysis Example: Compressive strength of a finished good can be controlled by 3 independent variables e.g. moisture in the material, grade of the raw material & pressure at which the product was produced. The values are shown below: Strength Moisture Content Grade Pressure 85 2.20 50 80 89 2.10 50 80 89 2.10 54 85 89 2.00 54 85 75 2.49 40 64 70 2.60 40 64 92 1.89 55 88 70 2.55 40 65 95 1.95 55 89 84 2.15 50 79 84 2.20 50 80 80 2.20 48 77 75 2.40 45 72 92 1.95 55 89 97 1.89 56 90 95 1.90 56 90 Regression analysis with effect of multicollinearity: Regression Analysis: Strength versus Moisture Content, Grade, Pressure Analysis of Variance Source DF Adj SS Adj MS F-Value P-Value Regression 3 1118.42 372.808 92.21 0.000 F is significant, overall model significant Moisture Content 1 16.82 16.821 4.16 0.064 Grade 1 0.71 0.708 0.18 0.683 Pressure 1 2.53 2.530 0.63 0.444 Model Summary S R-sq R-sq(adj) R-sq(pred) 2.01068 95.84% 94.80% 92.06% high R² with high R-sq (adj) & strong predictibility Coefficients Term Coef SE Coef T-Value P-Value VIF Constant 113.7 50.5 2.25 0.044 Moisture Content -24.5 12.0 -2.04 0.064 30.49 Grade -0.47 1.12 -0.42 0.683 157.37 variable (s) non-significant (>0.05), high VIF Pressure 0.598 0.755 0.79 0.444 181.53 variable (s) non-significant (>0.05), high VIF Regression Equation Strength = 113.7 - 24.5 Moisture Content - 0.47 Grade + 0.598 Pressure As per the data, If higher grade is used, higher strength is obtained whereas the sign in the equation is reverse (-ve sign, -0.47xgrade). i.e. 1 unit increase in grade, strength will be reduced by 0.47. Regression analysis without the effect of multicollinearity: Regression Analysis: Strength versus Moisture Content, Grade, Pressure Stepwise Selection of Terms α to enter = 0.05, α to remove = 0.05 Analysis of Variance Source DF Adj SS Adj MS F-Value P-Value Regression 1 1113.68 1113.68 292.74 0.000 F significant, overall model is significant Moisture Content 1 1113.68 1113.68 292.74 0.000 Model Summary S R-sq R-sq(adj) R-sq(pred) 1.95047 95.44% 95.11% 94.18% Standard error is less than previous one (2.01) high R² with high R-sq (adj) & strong predictibility Coefficients Term Coef SE Coef T-Value P-Value VIF Constant 163.11 4.59 35.55 0.000 Moisture Content -36.12 2.11 -17.11 0.000 1.00 variable significant (<0.05), low VIF Regression Equation Strength = 163.11 - 36.12 Moisture Content