Single and multi-group evacuation in medium and maxi health
Transcript
Single and multi-group evacuation in medium and maxi health
Single and multi-group evacuation in medium and maxi health-care emergencies Edoardo Amaldi DEI, Politecnico di Milano joint work with Stefano Coniglio and Claudio Iuliano Game Theory Workshop @ Milano – May 13, 2011 Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 1 / 17 Outline 1 The problem: evacuation in medium and maxi emergencies 2 Mixed integer linear programming formulation 3 Computational results Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 2 / 17 DECEMbRIA project DECEMbRIA (DECisioni in EMergenza sanitaRIA) project funded by Lombardy Region (2008-10) involving Operations Center of 118 Emergency Medical Service, Ospedale Niguarda Ca’ Granda, Milano Politecnico di Milano, Univ. di Milano, Univ. del Piemonte Orientale Collaboration with Dr. Cristiano Cozzi, head of the EU of 118 Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 3 / 17 Medium and maxi emergencies Large accidents with many injured people (large intoxication, major car accident, building collapse, airplane crash,...) Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 4 / 17 Medium and maxi emergencies Large accidents with many injured people (large intoxication, major car accident, building collapse, airplane crash,...) Flow of events 1 establishment of an aid station on the emergency site 2 diagnosis: determination for each patient of severity color code (red, yellow, green) and of needed medical specialties 3 assignment of each patient to a hospital Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 4 / 17 The problem Given for each patient: color code and needed medical specialties (general reanimation, coronary, neuro-reanimation, operating room) for each hospital: distance, capacities (number of curable patients) and estimated competences w.r.t. color codes and medical specialties number of available ambulances select for each patient an ambulance and a hospital so as to optimize total travel distance balance the residual capacities quality of treatments Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 5 / 17 The problem Peculiarities of the assignment problem packing on ambulances 1 single red-code 1 yellow-code possibly with 1 green-code up to 3 green-code activation of internal emergency plan for massive injured people inflow (PEIMAF) to increase the color-code capacities Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 6 / 17 Previous work urban evacuation, e.g., [Zeng, Wang, and Kukreti, 2008] building evacuation (flow over time), e.g., [Hoppe and Tardos, 1994] military aeromedical evacuation, e.g., [Kott, Saks, and Mercer, 1999] The problem is currently tackled on a case-by-case basis, not addressed as an optimization problem Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 7 / 17 Scenarios Efficiently support decisions in the critical setting of a medium and maxi emergency single-group evacuation: all patients simultaneously Solving a mixed integer linear programming formulation multi-group evacuation: patients in groups (e.g., when extracted from the rubble after a building collapse) Solving the single-group formulation iteratively, avoiding early consumption of scarce resources Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 8 / 17 Parameters C = {r , y , g }: color codes (red, yellow, green) Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 9 / 17 Parameters C = {r , y , g }: color codes (red, yellow, green) S: medical specialties (general reanimation, coronary, neuro-reanimation, operating room) Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 9 / 17 Parameters C = {r , y , g }: color codes (red, yellow, green) S: medical specialties (general reanimation, coronary, neuro-reanimation, operating room) P: patients, with Sp ⊆ S needed specialties for each p ∈ P Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 9 / 17 Parameters C = {r , y , g }: color codes (red, yellow, green) S: medical specialties (general reanimation, coronary, neuro-reanimation, operating room) P: patients, with Sp ⊆ S needed specialties for each p ∈ P Pc and Ps ⊆ P: patients with color code c and requiring specialty s Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 9 / 17 Parameters C = {r , y , g }: color codes (red, yellow, green) S: medical specialties (general reanimation, coronary, neuro-reanimation, operating room) P: patients, with Sp ⊆ S needed specialties for each p ∈ P Pc and Ps ⊆ P: patients with color code c and requiring specialty s [ H= Hi : set of hospitals partitioned into geographic zones i ∈ G i∈G For each h ∈ H: dh ∈ R+ distance to emergency site 0 capacity khc and khs per color code and specialty 0 competences bhc and bhs in {0, . . . , 10} per color code and specialty extra capacity with PEIMAF k̂hc per color PEIMAF activation cost wh Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 9 / 17 Parameters C = {r , y , g }: color codes (red, yellow, green) S: medical specialties (general reanimation, coronary, neuro-reanimation, operating room) P: patients, with Sp ⊆ S needed specialties for each p ∈ P Pc and Ps ⊆ P: patients with color code c and requiring specialty s [ H= Hi : set of hospitals partitioned into geographic zones i ∈ G i∈G For each h ∈ H: dh ∈ R+ distance to emergency site 0 capacity khc and khs per color code and specialty 0 competences bhc and bhs in {0, . . . , 10} per color code and specialty extra capacity with PEIMAF k̂hc per color PEIMAF activation cost wh number of available ambulances a0 Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 9 / 17 Parameters C = {r , y , g }: color codes (red, yellow, green) S: medical specialties (general reanimation, coronary, neuro-reanimation, operating room) P: patients, with Sp ⊆ S needed specialties for each p ∈ P Pc and Ps ⊆ P: patients with color code c and requiring specialty s [ H= Hi : set of hospitals partitioned into geographic zones i ∈ G i∈G For each h ∈ H: dh ∈ R+ distance to emergency site 0 capacity khc and khs per color code and specialty 0 competences bhc and bhs in {0, . . . , 10} per color code and specialty extra capacity with PEIMAF k̂hc per color PEIMAF activation cost wh number of available ambulances a0 upper bound w0 on “cost” for activating PEIMAFs Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 9 / 17 Variables Decision variables xph ∈ {0, 1}: 1 if patient p is assigned to hospital h and 0 otherwise yh ∈ {0, 1}: 1 if PEIMAF is activated in h and 0 otherwise zh ∈ Z+ : number of ambulances sent to hospital h uc ∈ Z+ : minimum residual capacity for color code c over all zones us0 ∈ Z+ : minimum residual capacity for specialty s over all zones Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 10 / 17 Variables Decision variables xph ∈ {0, 1}: 1 if patient p is assigned to hospital h and 0 otherwise yh ∈ {0, 1}: 1 if PEIMAF is activated in h and 0 otherwise zh ∈ Z+ : number of ambulances sent to hospital h uc ∈ Z+ : minimum residual capacity for color code c over all zones us0 ∈ Z+ : minimum residual capacity for specialty s over all zones Auxiliary variables rhc ∈ [0, khc ]: residual capacity for color code c in hospital h 0 0 rhs ∈ [0, khs ]: residual capacity for specialty s in hospital h zh0 ∈ Z+ : packing on ambulances for hospital h Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 10 / 17 Formulation X min αc xph dh − − γc uc − c∈C c∈C ,p∈Pc ,h∈H X X γs0 us0 s∈S X βc xph bhc − c∈C ,p∈Pc ,h∈H X 0 βs0 xph bhs s∈S,p∈Ps ,h∈H s.t. X ∀p ∈ P xph = 1 (assignment) h∈H X wh yh ≤ w0 (PEIMAF limit) h∈H X xph + rhc ≤ khc + yh k̂hc ∀h ∈ H, c ∈ C (color residual) 0 0 xph + rhs = khs ∀h ∈ H, s ∈ S (specialty residual) p∈Pc X p∈Ps ... Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 11 / 17 Formulation X min αc xph dh − − γc uc − c∈C c∈C ,p∈Pc ,h∈H X X γs0 us0 s∈S X βc xph bhc − c∈C ,p∈Pc ,h∈H X 0 βs0 xph bhs s∈S,p∈Ps ,h∈H s.t. X ∀p ∈ P xph = 1 (assignment) h∈H X wh yh ≤ w0 (PEIMAF limit) h∈H X xph + rhc ≤ khc + yh k̂hc ∀h ∈ H, c ∈ C (color residual) 0 0 xph + rhs = khs ∀h ∈ H, s ∈ S (specialty residual) p∈Pc X p∈Ps ... Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 11 / 17 Formulation X min αc xph dh − − γc uc − c∈C c∈C ,p∈Pc ,h∈H X X γs0 us0 s∈S X βc xph bhc − c∈C ,p∈Pc ,h∈H X 0 βs0 xph bhs s∈S,p∈Ps ,h∈H s.t. X ∀p ∈ P xph = 1 (assignment) h∈H X wh yh ≤ w0 (PEIMAF limit) h∈H X xph + rhc ≤ khc + yh k̂hc ∀h ∈ H, c ∈ C (color residual) 0 0 xph + rhs = khs ∀h ∈ H, s ∈ S (specialty residual) p∈Pc X p∈Ps ... Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 11 / 17 Formulation X min αc xph dh − − γc uc − c∈C c∈C ,p∈Pc ,h∈H X X γs0 us0 s∈S X βc xph bhc − c∈C ,p∈Pc ,h∈H X 0 βs0 xph bhs s∈S,p∈Ps ,h∈H s.t. X ∀p ∈ P xph = 1 (assignment) h∈H X wh yh ≤ w0 (PEIMAF limit) h∈H X xph + rhc ≤ khc + yh k̂hc ∀h ∈ H, c ∈ C (color residual) 0 0 xph + rhs = khs ∀h ∈ H, s ∈ S (specialty residual) p∈Pc X p∈Ps ... Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 11 / 17 Formulation X min αc xph dh − γc uc − X γs0 us0 X βc xph bhc − c∈C ,p∈Pc ,h∈H X s∈S c∈C c∈C ,p∈Pc ,h∈H − X 0 βs0 xph bhs s∈S,p∈Ps ,h∈H s.t. uc ≤ X rhc ∀c ∈ C , i ∈ G (min res. color) 0 rhs ∀s ∈ S, i ∈ G (min res. specialty) h∈Hi us0 ≤ X h∈Hi X z h ≤ a0 (available ambulances) h∈H zh = X xph + zh0 ∀h ∈ H (amb. number) ∀h ∈ H (amb. packing) p∈(Pr ∪Py ) P zh0 ≥ p∈Pg xph − Edoardo Amaldi (Politecnico di Milano) 3 P p∈Py xph Single and multi-group evacuation in medium and maxi health-care emergencies 12 / 17 Formulation X min αc xph dh − γc uc − X γs0 us0 X βc xph bhc − c∈C ,p∈Pc ,h∈H X s∈S c∈C c∈C ,p∈Pc ,h∈H − X 0 βs0 xph bhs s∈S,p∈Ps ,h∈H s.t. uc ≤ X rhc ∀c ∈ C , i ∈ G (min res. color) 0 rhs ∀s ∈ S, i ∈ G (min res. specialty) h∈Hi us0 ≤ X h∈Hi X z h ≤ a0 (available ambulances) h∈H zh = X xph + zh0 ∀h ∈ H (amb. number) ∀h ∈ H (amb. packing) p∈(Pr ∪Py ) P zh0 ≥ p∈Pg xph − Edoardo Amaldi (Politecnico di Milano) 3 P p∈Py xph Single and multi-group evacuation in medium and maxi health-care emergencies 12 / 17 Formulation X min αc xph dh − γc uc − X γs0 us0 X βc xph bhc − c∈C ,p∈Pc ,h∈H X s∈S c∈C c∈C ,p∈Pc ,h∈H − X 0 βs0 xph bhs s∈S,p∈Ps ,h∈H s.t. uc ≤ X rhc ∀c ∈ C , i ∈ G (min res. color) 0 rhs ∀s ∈ S, i ∈ G (min res. specialty) h∈Hi us0 ≤ X h∈Hi X z h ≤ a0 (available ambulances) h∈H zh = X xph + zh0 ∀h ∈ H (amb. number) ∀h ∈ H (amb. packing) p∈(Pr ∪Py ) P zh0 ≥ p∈Pg xph − Edoardo Amaldi (Politecnico di Milano) 3 P p∈Py xph Single and multi-group evacuation in medium and maxi health-care emergencies 12 / 17 Formulation X min αc xph dh − γc uc − X γs0 us0 X βc xph bhc − c∈C ,p∈Pc ,h∈H X s∈S c∈C c∈C ,p∈Pc ,h∈H − X 0 βs0 xph bhs s∈S,p∈Ps ,h∈H s.t. uc ≤ X rhc ∀c ∈ C , i ∈ G (min res. color) 0 rhs ∀s ∈ S, i ∈ G (min res. specialty) h∈Hi us0 ≤ X h∈Hi X z h ≤ a0 (available ambulances) h∈H zh = X xph + zh0 ∀h ∈ H (amb. number) ∀h ∈ H (amb. packing) p∈(Pr ∪Py ) P zh0 ≥ p∈Pg xph − Edoardo Amaldi (Politecnico di Milano) 3 P p∈Py xph Single and multi-group evacuation in medium and maxi health-care emergencies 12 / 17 Characteristics Optimal patient-to-ambulance assignment implicitly accounted for by the last constraints explicitly derived by a greedy procedure after solving the model Proposition: Integrality of the solution guaranteed even with continuous assignment variables xph Extension to multiple simultaneous emergencies at different sites Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 13 / 17 Computational results Hard instances: 10 − 300 patients random color-code in {r , y , g } (respectively 12%, 25% and 63%) unique medical specialty small groups in multi-group scenario (5 patients) minimum number of ambulances in both scenarios regular and PEIMAF color-code capacities, specialty capacities and competences based on the experience of the EU operators Milan and hinterland subdivided into 10 geographic zones Averages on 30 instances, MIP formulation(s) solved via CPLEX 11 Intel Xeon 2.0 GHz, with 4 GB RAM (1 core and RAM usage < 1 GB) Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 14 / 17 Computational results |P| 10 15 20 25 30 35 40 45 50 60 70 80 90 100 120 150 180 200 250 300 Edoardo Amaldi (Politecnico di Milano) single-group |A| Time [s] 6 0.02 9 0.03 11 0.03 13 0.03 16 0.04 19 0.05 21 0.08 24 0.05 26 0.06 31 0.07 36 0.14 41 0.13 46 0.23 51 0.23 60 0.29 75 0.54 91 0.98 101 1.18 127 1.86 150 4.15 multi-group |A| Time [s] 6 0.04 9 0.05 11 0.07 14 0.09 16 0.11 19 0.13 21 0.16 24 0.18 27 0.20 31 0.27 37 0.31 42 0.37 47 0.43 52 0.53 62 0.66 77 1.08 94 1.26 106 1.66 133 2.09 157 3.50 Gap [%] 0.00 0.00 0.00 1.93 0.20 0.47 0.81 0.96 0.97 4.83 3.15 3.10 3.36 2.87 2.88 3.10 2.49 2.62 2.82 1.75 Single and multi-group evacuation in medium and maxi health-care emergencies 15 / 17 Graphical user interface Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 16 / 17 Concluding remarks We propose a mixed integer linear formulation for the single-group evacuation an iterative approach for the multi-group scenario CPU time required to solve our formulations is sufficiently small to suit the limited time available in critical emergency situations Need for up-to-date and accurate data from the hospitals to use our optimization tool Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 17 / 17 Concluding remarks We propose a mixed integer linear formulation for the single-group evacuation an iterative approach for the multi-group scenario CPU time required to solve our formulations is sufficiently small to suit the limited time available in critical emergency situations Need for up-to-date and accurate data from the hospitals to use our optimization tool Edoardo Amaldi (Politecnico di Milano) Single and multi-group evacuation in medium and maxi health-care emergencies 17 / 17