Reorder Levels and Safety Stock When Lead Times Are Erratic

In most hospital consumable categories, consumption is comparatively stable and supply is not. Daily glove usage moves within a fairly narrow band. Distributor delivery against the same order might take four days one month and nineteen the next, depending on stock at the depot, a batch under test, a transporter shutdown or a payment dispute nobody told you about. Safety stock built only on demand variation will be systematically too small.

The formula that handles both is a combination: safety stock scales with the square root of demand variance across the lead time plus the variance of the lead time itself multiplied by average demand. The algebra matters less than the intuition it encodes, which is that lengthening and destabilising the lead time raises your required buffer faster than the same proportional increase in demand variability does. It is why a supplier who is slow but predictable is cheaper to hold than one who is fast on average and occasionally disappears.

This gives you a negotiating argument that materials teams rarely make. A supplier's lead-time consistency has a measurable inventory cost, and you can quantify it. Where two suppliers quote similar rates and one has visibly tighter delivery performance, the difference in safety stock they oblige you to carry is a real number you can put in the comparative statement.

The standard safety stock formula assumes demand is roughly normally distributed. For fast-moving ward consumables that assumption is tolerable. For an item consumed twice a month in a single procedure it is nonsense, and applying the formula produces a level that is either absurd or zero. Before you compute anything, split the item master by demand pattern rather than by department.

For steady movers, use the statistical approach and let the service level do the work. For lumpy or intermittent items, ignore the formula and set the level on a rule instead: enough to cover the largest single procedure requirement plus one lead time, or a fixed number of procedure sets. For genuinely one-off items, do not hold stock at all; buy against the confirmed case and manage the lead time clinically by scheduling around it.

Then sanity-check every calculated level against days of cover and against shelf life. A number that survives the arithmetic but implies eleven months of cover on a nine-month product has failed. Build that check into the review as a hard rule so nobody has to remember it, and route the failures to a human rather than letting the system silently truncate them.

The service level you target is a policy choice about how often you are willing to stock out, and it should not be uniform. Applying a high service level to every item is expensive and fills your store with slow-moving cover. Applying a moderate one to everything means periodically running out of something that matters clinically. The criticality classification your pharmacy and therapeutics committee already maintains is the natural input.

Set two or three tiers. Vital items where a stockout has an immediate clinical consequence get the highest service level and, where the supply base allows, a second approved source. Essential items get a moderate level. Desirable items get a modest one and are allowed to stock out occasionally, because the cost of never running out of a substitutable item exceeds the cost of running out of it.

Say the trade-off out loud when you set it, because otherwise the first stockout in the lowest tier will be treated as a failure of the system rather than as the designed behaviour it is. Write the intended stockout tolerance into the inventory policy, get it approved by the same committee that classified the items, and report stockouts against tier so the conversation stays about whether the tiering is right.

A statistical model built on twelve months of average behaviour will fail predictably at the points in the year when the supply chain does not behave averagely. The major festival periods shut down transport and distributor operations for several days at a time, and the timing varies by state and by the specific festivals your suppliers observe. Financial year end brings its own distortion, with distributors managing their own closing stock and sometimes declining to supply.

Monsoon disrupts road logistics in parts of the country for weeks, and cold-chain items suffer disproportionately because a delayed shipment is not merely late but may be unusable. Add to that the unscheduled events: a manufacturing site under regulatory action, a batch held for testing, a recall pulling a substitute product into shortage across the whole market.

Handle these with a calendar overlay rather than by inflating the base level year round. Build an annual supply calendar marking the periods where lead times extend, raise levels for the affected items a lead time before each window, and drop them afterwards. It is more work than a static level and considerably cheaper than carrying festival cover for twelve months.

Some categories will defeat any level-setting approach and need a different mechanism entirely. Consignment implants, where the stock physically sits with you but is owned by the supplier until used, do not have a reorder point in the conventional sense; they have a replenishment agreement and a reconciliation discipline. Reagents tied to an analyser follow test volume, which follows clinical demand and machine uptime rather than historical issue patterns.

Narcotic and psychotropic items are governed by their own custody, register and licensing requirements, and the constraint on holding is regulatory rather than economic. Blood products are not stocked in any ordinary sense. High-value items where holding cost is genuinely material may be better served by a supplier commitment to deliver within hours than by carrying cover, if the supply base makes that credible.

The failure mode to avoid is applying the general formula to these categories and then quietly overriding it every month. Overrides that happen every cycle are not exceptions; they are evidence that the rule is wrong for that category. Pull those items out of the standard model, give them a named replenishment rule, and record what the rule is so the next materials manager does not have to rediscover it.

Levels decay. Case mix shifts, a new consultant changes a preferred consumable, a department opens, a supplier is replaced, and the parameters set eighteen months ago now describe a hospital that no longer exists. Review them on a cycle — quarterly for the fast movers and the vital items, annually for the rest — and treat the review as a data exercise rather than a discussion.

The three reports that drive the review are stockout events by item and tier, items sitting above maximum for more than a stated number of days, and items whose actual lead time has drifted from the parameter used to calculate their level. The third is the one most often missing and the most predictive, because a supplier whose lead time has quietly doubled will cause a stockout before any other report notices.

Give the review an output that changes the system, not a note in the minutes. Parameters updated in the inventory master with the date and the reason, items reclassified between tiers with clinical sign-off, and any item moved out of the standard model recorded with its new rule. An inventory system holding consumption, lead time and level history against the same item makes this a two-hour exercise; without that, it is a project nobody schedules.