Designing a Hospital Diabetes and Chronic Care Programme

The registry is not a spreadsheet of patient names. It is a per-patient record of the small set of facts the programme acts on, each with a date and a due date: the last HbA1c, the last blood pressure, the last lipid profile, the last renal function and urine albumin, the last retinal examination, the last foot examination, current medications, and the date of the next scheduled contact. Everything else lives in the clinical record. The registry holds only what drives an action.

The reason to keep it that tight is that the registry's job is to produce work lists. Who is overdue for HbA1c. Who has not had a retinal screen this year. Whose last reading was above target and who has not been seen since. Those queries have to run in seconds and be actionable by a coordinator without clinical training, and every additional field you add is a field somebody has to maintain and nobody will.

Data quality will be the first shock. Most hospitals discover that a large share of their diabetic patients have no coded diagnosis, that blood pressure is recorded in free text, and that a laboratory result exists as a PDF rather than a value. Fixing this is unglamorous and it is the actual project. Structured vitals capture at the OPD, coded problem lists, and discrete laboratory results are prerequisites, not enhancements, and any plan that assumes they already work will stall in month two.

Recall is a scheduled, tracked outreach cycle, and it is where most programmes quietly stop. The mechanism is simple: the registry generates a due list, a coordinator works it, each contact attempt is recorded with its outcome, and a patient moves to a different track after a defined number of failed attempts. What makes it hard is that it is repetitive, low-status work that produces no visible result on the day it is done, so it is the first thing dropped when the coordinator is pulled to cover the front desk.

Channel choice should follow the patient, not the hospital's preference. A phone call from a person the patient recognises works better than any message for elderly patients and for those who do not read comfortably. A message in the patient's registered language works for younger working patients who will not take a call. Whatever you use, consent and preference need to be recorded per channel, and the rules for commercial communication and template approval on messaging platforms apply here just as they do to appointment reminders.

Adherence tracking is harder than it looks because dispensing is not the same as taking. Pharmacy refill data from your own outpatient pharmacy is the most reliable signal you can get cheaply: a patient collecting a month's supply every forty-five days is not taking it as prescribed. That is a genuinely useful flag and it requires no new data collection at all, only that the outpatient pharmacy record links to the same patient identity as the clinic.

The screening burden in diabetes care is well established: periodic retinal examination, annual foot risk assessment, renal function and albuminuria, lipids, and cardiovascular risk review, at intervals set by your clinical protocol. The programme's job is not to decide the intervals, which come from your clinicians and national guidance including the framework under the national NCD programme. Its job is to make sure that the interval is tracked, the appointment is made, and the result comes back to the record.

Assign each screen an owner and a route. Retinal screening is the one that most often fails, because it needs an ophthalmology slot the diabetes clinic does not control, and patients sent to book it themselves largely do not. The programmes that succeed either run a fundus camera in the diabetes clinic with reporting by an ophthalmologist, or hold protected ophthalmology slots the coordinator can book directly. Both cost something. Neither costs as much as a cohort with no retinal surveillance.

Foot examination is the opposite problem: it is cheap, takes ten minutes, needs no equipment beyond a monofilament, and is skipped because nobody's name is against it. Make it a nurse-delivered structured assessment with a risk category as the output, recorded as discrete fields, and audit completion monthly. A programme that can show a completion rate for each screen is managing something. One that cannot is hoping.

A monthly multidisciplinary meeting with the physician, a diabetes educator, a dietitian, a pharmacist, and the programme coordinator is where the registry turns into decisions. It only works if the agenda is generated by the registry rather than by whoever remembers a difficult patient. Bring the three lists that matter: patients above target despite maximal current therapy, patients with a new complication flag, and patients lost to follow-up after repeated contact attempts.

Keep the meeting short and make the output a set of assigned actions with dates in the registry, not minutes in a document. If the pharmacist agrees to review a patient's regimen, that becomes a task with a due date attached to the patient record, and it appears on next month's agenda if it is not done. Without this loop the meeting becomes a pleasant clinical discussion that changes nothing, which is what most such meetings are.

The staffing reality deserves honesty. Diabetes educators and dietitians are the roles hospitals cut first and they are the roles that deliver most of the behaviour change the programme depends on. A programme funded only for a physician clinic and a coordinator will produce good measurement and mediocre outcomes. That may still be the right starting point, but it should be a stated choice rather than an accident of the budget.

Chronic care programmes die of funding, not of clinical failure, because the work that creates the value is the work nobody bills for. Recall calls, registry maintenance, and educator time have no consultation attached. Meanwhile the value the programme creates, in retained patients, laboratory volume, pharmacy revenue, and avoided complications, lands in other departments' numbers. Unless someone makes that flow visible, the programme reads as a cost centre and is cut in the first difficult quarter.

There are three workable models and most hospitals end up blending them. An annual care package with a defined set of reviews, screens, and tests sold as a bundle gives the programme its own revenue line and gives patients a clear price. A cross-subsidy model funds the coordinator from the attributable diagnostic and pharmacy margin, which requires attribution reporting that actually works. And for insured or scheme-covered patients, structuring reviews as billable consultations with the ancillary tests coded correctly recovers part of the cost.

Whichever you pick, build the attribution reporting before you need to defend the budget. Show, month by month, the laboratory revenue, pharmacy revenue, and consultation revenue generated by enrolled patients against the direct programme cost. HealUDoc reporting can attribute downstream activity to the enrolling programme so the contribution is visible rather than argued. Present it as contribution rather than profit, because that is what it honestly is, and an overclaimed business case is worse than a modest one.

The measures that describe a programme are population measures, and they look different from clinic measures. What proportion of the panel has an HbA1c recorded in the last six months. What proportion is at the agreed target. What proportion has had each due screen. What proportion has been contacted in the last quarter and what proportion is genuinely lost. Report all of these with the panel as the denominator, including the patients who never came, because those are precisely the patients a clinic-based measure makes invisible.

Expect the first set of numbers to be bad and plan the communication accordingly. A programme that reports sixty per cent screening coverage in its first quarter has not performed badly; it has measured honestly something that was previously unmeasured and probably worse. The risk is that a management committee reads the first report as failure and withdraws support before the trend exists. Say clearly, in advance, that the first two quarters establish a baseline.

Finally, review the panel definition annually. Patients die, move, transfer their care, and recover, and a registry that only ever grows becomes a list nobody trusts and nobody works. A defined exit process, with a recorded reason for each removal, keeps the denominator honest. It also keeps the coordinator's work list finite, which is the practical difference between a programme that runs for years and one that collapses under its own accumulated backlog.